summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorLance Martin <lance@anthropic.com>2026-09-22 10:13:51 -0700
committerLance Martin <lance@anthropic.com>2026-09-22 10:13:51 -0700
commit1e242283a117bd9f638160a8e7789362ef4528aa (patch)
tree7cc3ec6b8ac07cdc12b1afbabf2ad3769fa205bb
parent22442b13357d7d0ddc166aeb6bbb7593cd761d35 (diff)
Update claude-api skill: Claude Opus 5.5 migration section and API changesrlm/claude-api-opus-5-5
- Add "Migrating to Claude Opus 5.5" to shared/model-migration.md: thinking can't be disabled (effort is the only control, default `medium`), forced `tool_choice` returns 400, thinking blocks tied to the model and the conversation, computer use only through `computer_toolset_20260801`; progress-update thinking blocks, safeguards, pricing, and a checklist - Wire Claude Opus 5.5 (`claude-opus-5-5`) into SKILL.md, models.md, error-codes.md, prompt-caching.md, prompt-audit.md, and platform-availability.md - Document the computer toolset request shape in tool-use-concepts.md
-rw-r--r--skills/claude-api/SKILL.md28
-rw-r--r--skills/claude-api/shared/error-codes.md11
-rw-r--r--skills/claude-api/shared/live-sources.md2
-rw-r--r--skills/claude-api/shared/model-migration.md229
-rw-r--r--skills/claude-api/shared/models.md5
-rw-r--r--skills/claude-api/shared/platform-availability.md10
-rw-r--r--skills/claude-api/shared/prompt-audit.md19
-rw-r--r--skills/claude-api/shared/prompt-caching.md4
-rw-r--r--skills/claude-api/shared/tool-use-concepts.md6
9 files changed, 274 insertions, 40 deletions
diff --git a/skills/claude-api/SKILL.md b/skills/claude-api/SKILL.md
index 2de6bed..8af5ca5 100644
--- a/skills/claude-api/SKILL.md
+++ b/skills/claude-api/SKILL.md
@@ -182,6 +182,7 @@ Everything goes through `POST /v1/messages`. Tools and output constraints are fe
| Claude Fable 5.1 | `claude-fable-5-1` | 1M | $10.00 | $50.00 |
| Claude Mythos 5.1 (Project Glasswing only) | `claude-mythos-5-1` | 1M | $10.00 | $50.00 |
| Claude Fable 5 | `claude-fable-5` | 1M | $10.00 | $50.00 |
+| Claude Opus 5.5 (launching - use only when the user names it) | `claude-opus-5-5` | 1M | $4.00 | $20.00 |
| Claude Opus 5 | `claude-opus-5` | 1M | $5.00 | $25.00 |
| Claude Opus 4.8 | `claude-opus-4-8` | 1M | $5.00 | $25.00 |
| Claude Opus 4.7 | `claude-opus-4-7` | 1M | $5.00 | $25.00 |
@@ -205,7 +206,11 @@ Claude Fable 5.1 is Anthropic's most capable widely released model, for the most
- **No assistant prefill** - same as the rest of the 4.6+ family.
- **30-day data retention required** - Claude Fable 5.1 is not available under zero data retention unless expressly authorized by Anthropic; requests from an org whose retention configuration doesn't meet the requirement return `400 invalid_request_error`.
- **Longer turns, different prompting** - single requests on hard tasks can run many minutes (plan timeouts/streaming/progress UX); effort sweeps should include low/medium for routine work; prompts written for prior models are often too prescriptive and reduce output quality. See `shared/model-migration.md` -> Migrating to Claude Fable 5.1 -> Behavioral shifts (prompt-tunable) for the recommended prompt snippets.
-- **Successor to Claude Fable 5 (`claude-fable-5`, still served) in the same tier at the same per-token price.** Same surface as Claude Fable 5 with three breaking changes - forced tool use (`tool_choice` `any` / `tool`) returns a 400 (use `auto` + a prompt instruction, `strict: true` for schema-valid arguments, or structured outputs); thinking blocks are bound to the producing model (other models drop them, unbilled); and editing earlier turns invalidates thinking blocks ("preserved thinking"; new accounts created on/after 2026-08-31 get a 400 on edited history; later models enforce it for everyone - make every harness append-only and run the three-step check; the opt-in controls are per-platform, see `shared/platform-availability.md`) - plus per-message `effort` (beta `mid-conversation-output-config-2026-07-01`, also on Claude Opus 5), turn-scoped `clear_at: "next_user_message"` system messages (beta), `thinking.display: "updates"` progress notes (beta, all platforms), cache reads at $0.25/MTok (whether Claude Mythos 5.1 shares that rate is open at launch), and content provenance. Covered Model - ZDR orgs get `400 invalid_request_error` as on Claude Fable 5 (ZDR only if expressly authorized by Anthropic); no Priority Tier. Same tokenizer as Claude Fable 5. See `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5.
+- **Successor to Claude Fable 5 (`claude-fable-5`, still served) in the same tier at the same per-token price.** Same surface as Claude Fable 5 with three breaking changes - forced tool use (`tool_choice` `any` / `tool`) returns a 400 (use `auto` + a prompt instruction, `strict: true` for schema-valid arguments, or structured outputs); thinking blocks are bound to the producing model (other models drop them, unbilled); and editing earlier turns invalidates thinking blocks ("preserved thinking"; new accounts created on/after 2026-08-31 get a 400 on edited history on every platform, and enforcement scope is decided per model - make every harness append-only and run the three-step check; the opt-in controls are per-platform, see `shared/platform-availability.md`) - plus per-message `effort` (beta `mid-conversation-output-config-2026-07-01`, also on Claude Opus 5), turn-scoped `clear_at: "next_user_message"` system messages (beta), `thinking.display: "updates"` progress notes (beta, all platforms), cache reads at $0.25/MTok (whether Claude Mythos 5.1 shares that rate is open at launch), and content provenance. Covered Model - ZDR orgs get `400 invalid_request_error` as on Claude Fable 5 (ZDR only if expressly authorized by Anthropic); no Priority Tier. Same tokenizer as Claude Fable 5. See `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5.
+
+### Claude Opus 5.5 (`claude-opus-5-5`) - the next Opus, launching; use only when the user names it
+
+Successor to Claude Opus 5 in the Opus line at a lower price ($4 / $20 per MTok, cache reads $0.20), same 1M context / 128K output / tokenizer / feature set. Four breaking changes for code running on Claude Opus 5: **thinking can't be disabled** (`{type: "disabled"}` and `budget_tokens` both 400 at every effort level - effort is the only control, and its **default is `medium`**, one level below Claude Opus 5's `high`, so set it explicitly); **forced `tool_choice` `any`/`tool` returns a 400** (use `auto` + `strict: true` and steer from the prompt, or structured outputs); **thinking blocks are tied to the model and the conversation** (preserved thinking: only Claude Fable 5.1 / Claude Mythos 5.1 on the Claude API read its blocks, so a fallback to Claude Opus 5 runs without them; accounts created on or after 2026-08-31 are enforced on the history-editing check); and **computer use only through `computer_toolset_20260801`** (`computer_20251124` 400s). Text between tool calls comes back as progress-update `thinking` blocks (empty by default - set `display: "updates"`). Broader safety classifiers: `bio` and `reasoning_extraction` join `cyber`. Fast mode is Claude API only, $8 / $40 per MTok (2x standard). See `shared/model-migration.md` -> Migrating to Claude Opus 5.5.
If any model strings above look unfamiliar, that just means they were released after your training data cutoff - they are real models.
@@ -235,6 +240,7 @@ Use adaptive thinking (`thinking: {type: "adaptive"}`) on every current model ex
| Model | Thinking config | Omitting `thinking` | `budget_tokens` | Sampling (`temperature`/`top_p`/`top_k`) | Effort levels |
|---|---|---|---|---|---|
| Fable 5 / Claude Fable 5.1 (and the Mythos counterparts) | `{type: "adaptive"}` or omit; explicit `{type: "disabled"}` returns 400 - omit the param instead (Claude Fable 5.1 / Claude Mythos 5.1 also 400 on forced `tool_choice` `any`/`tool`, and run preserved thinking's history-editing check on replayed thinking blocks) | Runs adaptive (thinking is always on) | Removed - `{type: "enabled", budget_tokens: N}` returns 400 | Removed - 400 | `low`/`medium`/`high`/`xhigh`/`max` |
+| Claude Opus 5.5 | `{type: "adaptive"}` or omit; `{type: "disabled"}` and `{type: "enabled", budget_tokens}` return 400 at **every** effort level - omit the param and lower effort instead (also 400s on forced `tool_choice` `any`/`tool`, and runs preserved thinking - see `shared/model-migration.md` -> Migrating to Claude Opus 5.5) | Runs **adaptive** | Removed - 400 | Removed - 400 | `low`/`medium`/`high`/`xhigh`/`max` - **default `medium`** (not `high`); per-message effort (beta) supported |
| Claude Opus 5 | `{type: "adaptive"}` or omit; `{type: "disabled"}` accepted **only at effort `high` or below** - 400 at `xhigh`/`max`, and see the disabled-thinking pitfall below | Runs **adaptive** (thinking is on by default - unlike Opus 4.8/4.7) | Removed - 400 | Removed - 400 | `low`-`max` (all five) |
| Opus 4.8 / 4.7 | `{type: "adaptive"}` is the only on-mode; `{type: "disabled"}` accepted | Runs **without** thinking - set `{type: "adaptive"}` explicitly | Removed - 400 | Removed - 400 | `low`/`medium`/`high`/`xhigh`/`max` |
| Sonnet 5 | `{type: "adaptive"}` is the only on-mode; `{type: "disabled"}` accepted | Runs adaptive | Removed - 400 | Removed - 400 | `low`/`medium`/`high`/`xhigh`/`max` |
@@ -276,7 +282,7 @@ For placement patterns, architectural guidance, and the silent-invalidator audit
## Fast Mode (Quick Reference)
-**Research preview, Claude Opus 5 / Opus 4.8 only** - Claude API and Managed Agents, not Bedrock / Google Cloud / Foundry. Opus 4.7 fast mode has been removed: `speed: "fast"` on 4.7 returns an error. Fast mode on Claude Opus 5 is priced at $10 / $50 per MTok. Fast mode runs the same model at up to 2.5x higher output tokens per second, at premium pricing. Three things are required on every request: use the **beta** messages endpoint (`client.beta.messages....`), pass the beta flag `fast-mode-2026-02-01`, and set `speed: "fast"` as a top-level request parameter (not a header, not in `extra_body`).
+**Research preview, Claude Opus 5 / Claude Opus 5.5 / Opus 4.8 only** - Claude API and Managed Agents, not Bedrock / Google Cloud / Foundry. Opus 4.7 fast mode has been removed: `speed: "fast"` on 4.7 returns an error. Fast mode on Claude Opus 5 is priced at $10 / $50 per MTok; on Claude Opus 5.5, $8 / $40 (its fast-mode docs flip after the model launch - confirm before quoting). Fast mode runs the same model at up to 2.5x higher output tokens per second, at premium pricing. Three things are required on every request: use the **beta** messages endpoint (`client.beta.messages....`), pass the beta flag `fast-mode-2026-02-01`, and set `speed: "fast"` as a top-level request parameter (not a header, not in `extra_body`).
```python
client.beta.messages.create(
@@ -304,7 +310,7 @@ client.beta.messages.create(
## Task Budgets (Quick Reference)
-**Beta, Claude Opus 5 / Fable 5 / Claude Fable 5.1 (confirm at launch) / Sonnet 5 / Opus 4.8 / 4.7.** A task budget gives Claude a token ceiling for an agentic loop so it paces itself and finishes gracefully instead of being cut off - distinct from `max_tokens`, which is an enforced per-response ceiling the model is not aware of. Minimum `total`: 20,000. Set `task_budget` inside `output_config` on `client.beta.messages.stream(...)` with beta flag `task-budgets-2026-03-13` - use streaming so the large `max_tokens` doesn't hit HTTP timeouts (full details: `shared/model-migration.md` -> Task Budgets):
+**Beta, Claude Opus 5 / Claude Opus 5.5 / Fable 5 / Claude Fable 5.1 (confirm at launch) / Sonnet 5 / Opus 4.8 / 4.7.** A task budget gives Claude a token ceiling for an agentic loop so it paces itself and finishes gracefully instead of being cut off - distinct from `max_tokens`, which is an enforced per-response ceiling the model is not aware of. Minimum `total`: 20,000. Set `task_budget` inside `output_config` on `client.beta.messages.stream(...)` with beta flag `task-budgets-2026-03-13` - use streaming so the large `max_tokens` doesn't hit HTTP timeouts (full details: `shared/model-migration.md` -> Task Budgets):
```python
with client.beta.messages.stream(
@@ -493,7 +499,7 @@ The Quick Task Reference below uses the `{lang}/claude-api/FILE.md` path notatio
**Long-running conversations (may exceed context window):**
-> Read `{lang}/claude-api/README.md` - see Compaction section
-**Migrating to a newer model (Fable 5.1 / Fable 5 / Opus 5 / Opus 4.8 / Opus 4.7 / Opus 4.6 / Sonnet 5 / Sonnet 4.6), replacing a retired model, or translating `budget_tokens` / prefill patterns to the current API:**
+**Migrating to a newer model (Opus 5.5 / Fable 5.1 / Fable 5 / Opus 5 / Opus 4.8 / Opus 4.7 / Opus 4.6 / Sonnet 5 / Sonnet 4.6), replacing a retired model, or translating `budget_tokens` / prefill patterns to the current API:**
-> Read `shared/model-migration.md`
**Upgrading the Anthropic SDK package itself across a major version (`anthropic` 0.x -> 1.x: `httpx2`, awaited async `.with_raw_response`, removed deprecated parameters / aliases / Text Completions, Python >= 3.10) - or writing new code against a project already on 1.x:**
-> Read `{lang}/claude-api/sdk-upgrade.md` (currently Python only; other SDKs have no bundled major-version guide yet - use that SDK's CHANGELOG via `shared/live-sources.md`)
@@ -505,6 +511,8 @@ The Quick Task Reference below uses the `{lang}/claude-api/FILE.md` path notatio
-> Read `shared/evals/eval-hillclimb.md` - runs Step 0 -> Step 5 with a train/test split; test is scored every round and is the headline.
**Rendering an eval-hillclimb HTML report:**
-> Run `shared/evals/report/build-report.mjs` when it is on disk (EAP install), else `shared/evals/report/build-report-lite.mjs` (always extracted with this skill) - both consume the `_state.json` / `vN/` layout produced by the hillclimb guide and write the same `trajectory/scores.tsv`. Don't write a parallel one.
+**Migrating to, prompting, or tuning Claude Opus 5.5 (thinking can't be disabled, effort tuning and the `medium` default, forced tool use, computer toolset, progress updates, safeguard false positives, visual inputs / design outputs):**
+-> Read `shared/model-migration.md` -> Migrating to Claude Opus 5.5; the preserved-thinking mechanics it points at are under Migrating to Claude Fable 5.1 from Claude Fable 5
**Prompting or tuning Fable 5/5.1 (long turns, effort, verbosity, autonomous runs, sub-agents):**
-> Read `shared/model-migration.md` -> Migrating to Claude Fable 5.1 -> Behavioral shifts (prompt-tunable) + Long-running agent recommendations
**Prompting or tuning Claude Fable 5.1 (progress updates, parallel tool calls, writing density / formatting, autonomy, test sprawl, whole-file rewrites) or making a harness compatible with preserved thinking's history-editing check (history edits, compaction, per-turn reminders):**
@@ -557,13 +565,13 @@ Live documentation URLs are in `shared/live-sources.md`.
## Common Pitfalls
- Don't truncate inputs when passing files or content to the API. If the content is too long to fit in the context window, notify the user and discuss options (chunking, summarization, etc.) rather than silently truncating.
-- **Prefill removed (Fable 5, Claude Fable 5.1, Opus 5, Sonnet 5, and the 4.6/4.7/4.8 family):** Assistant message prefills (last-assistant-turn prefills) return a 400 error on Fable 5, Claude Fable 5.1, Opus 5, Sonnet 5, Opus 4.6, Opus 4.7, Opus 4.8, and Sonnet 4.6. Use structured outputs (`output_config.format`) or system prompt instructions to control response format instead. (One exception: the fallback-credit prefill claim - when redeeming a credit with `fallback_has_prefill_claim: true`, the server accepts the echoed assistant message; see the migration guide's refusal section.)
+- **Prefill removed (Fable 5, Claude Fable 5.1, Opus 5, Claude Opus 5.5, Sonnet 5, and the 4.6/4.7/4.8 family):** Assistant message prefills (last-assistant-turn prefills) return a 400 error on Fable 5, Claude Fable 5.1, Opus 5, Claude Opus 5.5, Sonnet 5, Opus 4.6, Opus 4.7, Opus 4.8, and Sonnet 4.6. Use structured outputs (`output_config.format`) or system prompt instructions to control response format instead. (One exception: the fallback-credit prefill claim - when redeeming a credit with `fallback_has_prefill_claim: true`, the server accepts the echoed assistant message; see the migration guide's refusal section.)
- **Confirm migration scope before editing:** When a user asks to migrate code to a newer Claude model without naming a specific file, directory, or file list, **ask which scope to apply first** - the entire working directory, a specific subdirectory, or a specific set of files. Do not start editing until the user confirms. Imperative phrasings like "migrate my codebase", "move my project to X", "upgrade to Sonnet 4.6", or bare "migrate to Opus 4.8" are **still ambiguous** - they tell you what to do but not where, so ask. Proceed without asking only when the prompt names an exact file, a specific directory, or an explicit file list ("migrate `app.py`", "migrate everything under `services/`", "update `a.py` and `b.py`"). See `shared/model-migration.md` Step 0.
- **`max_tokens` defaults:** Don't lowball `max_tokens` - hitting the cap truncates output mid-thought and requires a retry. For non-streaming requests, default to `~16000` (keeps responses under SDK HTTP timeouts). For streaming requests, default to `~64000` (timeouts aren't a concern, so give the model room). Only go lower when you have a hard reason: classification (`~256`), cost caps, deliberately short outputs, or **`max_tokens: 0`** for cache pre-warming (see `shared/prompt-caching.md` -> Pre-warming).
-- **Disabling thinking on Claude Opus 5 has two failure modes - prefer low/medium effort instead.** Only affects code that explicitly opts out; thinking is on by default, so watch for a disabled-thinking setting carried forward from Opus 4.8. With `thinking: {type: "disabled"}`, the model occasionally writes a tool call into its **visible text** instead of a `tool_use` block: the turn succeeds, the call never runs, no error is raised, and in an agentic loop that text pollutes later turns. It can also leak `<thinking>` tags into the response. Turning thinking on and lowering `effort` fixes both and still cuts cost. If a route must stay thinking-off: **delete** any don't-think/don't-reason rule (it makes tag leakage worse), don't name thinking tags, and add the combined instruction *"When you use a tool, you may say a brief sentence first. If no tool can express what the user asked for, say so instead of guessing. Do not include internal or system XML tags in your response."* Details: `shared/model-migration.md` -> Two failure modes when thinking is disabled.
-- **128K output tokens:** Fable 5, Claude Fable 5.1, Opus 5, Opus 4.6, Opus 4.7, Opus 4.8, Sonnet 5, and Sonnet 4.6 support up to 128K `max_tokens`, but the SDKs require streaming for values that large to avoid HTTP timeouts. Use `.stream()` with `.get_final_message()` / `.finalMessage()`.
-- **Forced tool use removed (Claude Fable 5.1 / Claude Mythos 5.1, as on Mythos Preview):** `tool_choice: {type: "any"}` and `{type: "tool", name: ...}` return a 400 (`tool_choice: type "tool" and "any" are not supported for this model.`), on `count_tokens` and Batches too. Use `{type: "auto"}` plus an explicit instruction naming the tool, `strict: true` on the tool to keep schema-valid arguments, or structured outputs (`output_config.format`) when the forced call only existed to get JSON back. `{type: "none"}` is unaffected; `disable_parallel_tool_use` still works with `auto` (at most one call).
-- **Tool call JSON parsing (Fable 5, Claude Fable 5.1, Opus 5, and the 4.6/4.7/4.8 family):** Fable 5, Claude Fable 5.1, Opus 5, Opus 4.6, Opus 4.7, Opus 4.8, and Sonnet 4.6 may produce different JSON string escaping in tool call `input` fields (e.g., Unicode or forward-slash escaping). Always parse tool inputs with `json.loads()` / `JSON.parse()` - never do raw string matching on the serialized input.
+- **Disabling thinking on Claude Opus 5 has two failure modes - prefer low/medium effort instead.** (On Claude Opus 5.5 it can't be disabled at all - `{type: "disabled"}` is a 400 at every effort level; use `low` effort.) Only affects code that explicitly opts out; thinking is on by default, so watch for a disabled-thinking setting carried forward from Opus 4.8. With `thinking: {type: "disabled"}`, the model occasionally writes a tool call into its **visible text** instead of a `tool_use` block: the turn succeeds, the call never runs, no error is raised, and in an agentic loop that text pollutes later turns. It can also leak `<thinking>` tags into the response. Turning thinking on and lowering `effort` fixes both and still cuts cost. If a route must stay thinking-off: **delete** any don't-think/don't-reason rule (it makes tag leakage worse), don't name thinking tags, and add the combined instruction *"When you use a tool, you may say a brief sentence first. If no tool can express what the user asked for, say so instead of guessing. Do not include internal or system XML tags in your response."* Details: `shared/model-migration.md` -> Two failure modes when thinking is disabled.
+- **128K output tokens:** Fable 5, Claude Fable 5.1, Opus 5, Claude Opus 5.5, Opus 4.6, Opus 4.7, Opus 4.8, Sonnet 5, and Sonnet 4.6 support up to 128K `max_tokens`, but the SDKs require streaming for values that large to avoid HTTP timeouts. Use `.stream()` with `.get_final_message()` / `.finalMessage()`.
+- **Forced tool use removed (Claude Fable 5.1 / Claude Mythos 5.1 / Claude Opus 5.5, as on Mythos Preview):** `tool_choice: {type: "any"}` and `{type: "tool", name: ...}` return a 400 (`tool_choice: type "tool" and "any" are not supported for this model.`), on `count_tokens` and Batches too. Use `{type: "auto"}` plus an explicit instruction naming the tool, `strict: true` on the tool to keep schema-valid arguments, or structured outputs (`output_config.format`) when the forced call only existed to get JSON back. `{type: "none"}` is unaffected; `disable_parallel_tool_use` still works with `auto` (at most one call).
+- **Tool call JSON parsing (Fable 5, Claude Fable 5.1, Opus 5, Claude Opus 5.5, and the 4.6/4.7/4.8 family):** Fable 5, Claude Fable 5.1, Opus 5, Claude Opus 5.5, Opus 4.6, Opus 4.7, Opus 4.8, and Sonnet 4.6 may produce different JSON string escaping in tool call `input` fields (e.g., Unicode or forward-slash escaping). Always parse tool inputs with `json.loads()` / `JSON.parse()` - never do raw string matching on the serialized input.
- **Structured outputs (all models):** Use `output_config: {format: {...}}` instead of the deprecated `output_format` parameter on `messages.create()`. This is a general API change, not 4.6-specific.
- **Don't reimplement SDK functionality:** The SDK provides high-level helpers - use them instead of building from scratch. Specifically: use `stream.finalMessage()` instead of wrapping `.on()` events in `new Promise()`; use typed exception classes (`Anthropic.RateLimitError`, etc.) instead of string-matching error messages; use SDK types (`Anthropic.MessageParam`, `Anthropic.Tool`, `Anthropic.Message`, etc.) instead of redefining equivalent interfaces.
- **Error handling - catch a chain, not one broad class.** A single `except APIStatusError` / `catch (AnthropicServiceException)` / `rescue APIError` loses the distinction between retryable (429, >=500, network) and non-retryable (400/404) failures. Write a most-specific-first chain - e.g. `NotFoundError` -> `RateLimitError` -> `APIStatusError` -> `APIConnectionError` (or the Go equivalent: `errors.As` into `*anthropic.Error` then `switch apierr.StatusCode { case 404: ...; case 429: ...; default: ... }`). Per-language class names and namespaces are in `shared/error-codes.md`.
@@ -576,7 +584,7 @@ Live documentation URLs are in `shared/live-sources.md`.
- **Fine-grained tool streaming is not a beta feature; this skill's default is to turn it on for streaming + client tools (the API itself still defaults to buffered).** Set `eager_input_streaming: true` on the tool definition and call the regular `client.messages.stream(...)`. There is no beta header and no `client.beta.*` path. Do not also send the legacy `fine-grained-tool-streaming-2025-05-14` beta header. Python's `@beta_tool(eager_input_streaming=True)` accepts it directly; TypeScript's `betaZodTool()` does not, so spread it on: `{ ...betaZodTool({...}), eager_input_streaming: true }`. With the field on, the API no longer coerces or validates the input, so the accumulated `partial_json` may be incomplete (`max_tokens`) or invalid - guard the parse (`shared/tool-use-concepts.md` -> Eager input streaming).
- **Cache diagnostics is beta.** Use `client.beta.messages.*` with beta `cache-diagnosis-2026-04-07`. Pass `diagnostics: {previous_message_id: null}` on the first turn and `diagnostics: {previous_message_id: <previous response id>}` on subsequent turns; the result is on `response.diagnostics`. Availability: `shared/platform-availability.md`.
- **Memory tool type is `memory_20250818`.** Declare `{"type": "memory_20250818", "name": "memory"}`. Go uses the beta-namespace type `{OfMemoryTool20250818: &anthropic.BetaMemoryTool20250818Param{}}` on `client.Beta.Messages.New`; Python/TypeScript/Ruby/PHP/C# use the non-beta `client.messages.create`; Java has both a non-beta `MemoryTool20250818` and a beta tool-runner path. Python/TypeScript provide `BetaAbstractMemoryTool` / `betaMemoryTool` helpers for implementing the backend.
-- **Use a model the feature actually supports.** Some features are restricted to specific model tiers - fast mode is Claude Opus 5 / Opus 4.8 only (and Claude API only), task budgets (Messages API only - Managed Agents session budgets have no model-tier restriction) are Claude Opus 5 / Fable 5 / Claude Fable 5.1 (confirm at launch) / Sonnet 5 / Opus 4.8 / 4.7 only, and the advisor tool requires a valid executor<->advisor pair. If the user's prompt names a model that the feature doesn't support, use a supported model instead and note the substitution in the output.
+- **Use a model the feature actually supports.** Some features are restricted to specific model tiers - fast mode is Claude Opus 5 / Claude Opus 5.5 / Opus 4.8 only (and Claude API only), task budgets (Messages API only - Managed Agents session budgets have no model-tier restriction) are Claude Opus 5 / Claude Opus 5.5 / Fable 5 / Claude Fable 5.1 (confirm at launch) / Sonnet 5 / Opus 4.8 / 4.7 only, and the advisor tool requires a valid executor<->advisor pair. If the user's prompt names a model that the feature doesn't support, use a supported model instead and note the substitution in the output.
- **Don't define custom types for SDK data structures:** The SDK exports types for all API objects. Use `Anthropic.MessageParam` for messages, `Anthropic.Tool` for tool definitions, `Anthropic.ToolUseBlock` / `Anthropic.ToolResultBlockParam` for tool results, `Anthropic.Message` for responses. Defining your own `interface ChatMessage { role: string; content: unknown }` duplicates what the SDK already provides and loses type safety.
- **Report and document output:** For tasks that produce reports, documents, or visualizations, the code execution sandbox has `python-docx`, `python-pptx`, `matplotlib`, `pillow`, and `pypdf` pre-installed. Claude can generate formatted files (DOCX, PDF, charts) and return them via the Files API - consider this for "report" or "document" type requests instead of plain stdout text.
- **Server-tool errors don't raise.** Web search and web fetch errors return HTTP 200 with a `web_search_tool_result` / `web_fetch_tool_result` block whose `content` is a single error object (e.g. `{error_code: "max_uses_exceeded"}`) - not a raised exception. For web search, a success `content` is a *list*; an error `content` is an *object* - branch on that before indexing.
diff --git a/skills/claude-api/shared/error-codes.md b/skills/claude-api/shared/error-codes.md
index 516b2dc..59a3762 100644
--- a/skills/claude-api/shared/error-codes.md
+++ b/skills/claude-api/shared/error-codes.md
@@ -121,8 +121,9 @@ Some 400 errors are specifically related to parameter validation:
- **Claude Opus 5:** `thinking: {type: "disabled"}` returns 400 when `effort` is `xhigh` or `max` - it is accepted at `high` or below. Thinking is on by default, so omitting the param runs adaptive rather than disabling it.
- **Fable 5/5.1 only:** an explicit `thinking: {type: "disabled"}` returns 400 at any effort (it is accepted on Opus 4.8/4.7). Omit the `thinking` param entirely instead.
- **Fable 5/5.1, Mythos 5/5.1:** if the organization or workspace is set to zero data retention (ZDR) - or any retention below the required 30 days - then **all** requests to these models return `400 invalid_request_error` ("In order to access this model, your organization or workspace must have data retention enabled."), even with a perfectly valid payload; ZDR only if expressly authorized by Anthropic. Check the retention configuration before debugging the request body.
-- **Claude Fable 5.1 / Claude Mythos 5.1 (and Mythos Preview):** `tool_choice: {type: "any"}` or `{type: "tool", name: ...}` returns 400 `tool_choice: type "tool" and "any" are not supported for this model.` - also on `count_tokens` and Batches. Use `{type: "auto"}` plus a prompt instruction (`strict: true` for schema-valid arguments), or structured outputs.
-- **Claude Fable 5.1 / Claude Mythos 5.1 - preserved thinking / history-editing check (new accounts created on/after 2026-08-31, or any request that sets `prefix_mismatch_behavior`):** ``messages.N.content.M: Invalid `signature` in `thinking` block. The block is bound to a different conversation. Remove the block, or set `thinking.block_binding.prefix_mismatch_behavior` to "drop_block".`` (plus a sentence naming the beta header when it wasn't sent, and optionally one naming the first message that changed) means the system prompt, tool list, or an earlier message changed since that thinking block was produced. Retrying the same body never clears it; `count_tokens` returns the same 400. (In the Message Batches API the *unset* default drops the failing blocks instead of failing the item - a Batches item fails as `errored` only with `prefix_mismatch_behavior: "error"` set.) Strip the named block and every thinking block after it and retry once, or resend with `thinking.block_binding.prefix_mismatch_behavior: "drop_block"` under beta `thinking-binding-controls-2026-08-01` (where the controls beta is offered - Claude API / Claude Platform on AWS at launch, per model on Bedrock and Google Cloud, not on Foundry: `shared/platform-availability.md`; elsewhere use the strip-and-retry path; without the header that field is a 400 ending `block_binding: Extra inputs are not permitted`); then fix the harness so it stops editing history (see `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5). The same leading clause with *no* "bound to a different conversation" sentence is a tampered signature - always a 400, regardless of the setting.
+- **Claude Opus 5.5:** `thinking: {type: "disabled"}` or `{type: "enabled", budget_tokens: N}` returns 400 `"thinking.type.disabled" is not supported for this model. Use "thinking.type.adaptive" and "output_config.effort" to control thinking behavior.` (`"thinking.type.enabled"` for the budget form) at every effort level - omit `thinking` and lower `output_config.effort` instead. A `tools` entry of type `computer_20251124` returns 400 `'claude-opus-5-5' does not support tool types: computer_20251124.` followed by `Did you mean one of` and the accepted types - declare `{type: "computer_toolset_20260801"}` instead (no beta header, no `name` / display size). See `shared/model-migration.md` -> Migrating to Claude Opus 5.5.
+- **Claude Fable 5.1 / Claude Mythos 5.1 / Claude Opus 5.5 (and Mythos Preview):** `tool_choice: {type: "any"}` or `{type: "tool", name: ...}` returns 400 `tool_choice: type "tool" and "any" are not supported for this model.` - also on `count_tokens` and Batches. Use `{type: "auto"}` plus a prompt instruction (`strict: true` for schema-valid arguments), or structured outputs.
+- **Claude Fable 5.1 / Claude Mythos 5.1 / Claude Opus 5.5 - preserved thinking / history-editing check (new accounts created on/after 2026-08-31 on every platform, or any request that sets `prefix_mismatch_behavior`):** ``messages.N.content.M: Invalid `signature` in `thinking` block. The block is bound to a different conversation. Remove the block, or set `thinking.block_binding.prefix_mismatch_behavior` to "drop_block".`` (plus a sentence naming the beta header when it wasn't sent, and optionally one naming the first message that changed) means the system prompt, tool list, or an earlier message changed since that thinking block was produced. Retrying the same body never clears it; `count_tokens` returns the same 400. (In the Message Batches API the *unset* default drops the failing blocks instead of failing the item - a Batches item fails as `errored` only with `prefix_mismatch_behavior: "error"` set.) Strip the named block and every thinking block after it and retry once, or resend with `thinking.block_binding.prefix_mismatch_behavior: "drop_block"` under beta `thinking-binding-controls-2026-08-01` (where the controls beta is offered - Claude API / Claude Platform on AWS at launch, per model on Bedrock and Google Cloud, not on Foundry: `shared/platform-availability.md`; elsewhere use the strip-and-retry path; without the header that field is a 400 ending `block_binding: Extra inputs are not permitted`); then fix the harness so it stops editing history (see `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5). The same leading clause with *no* "bound to a different conversation" sentence is a tampered signature - always a 400, regardless of the setting.
**Common mistake with extended thinking on older models (Opus 4.6 and earlier):**
@@ -184,8 +185,10 @@ thinking: budget_tokens=10000, max_tokens=16000
| `budget_tokens` on Claude Opus 5 / Fable 5/5.1 / Opus 4.8 / 4.7 | 400 | Use `thinking: {type: "adaptive"}` |
| `thinking: {type: "disabled"}` on Fable 5/5.1 | 400 | Omit the `thinking` param entirely (accepted on Opus 4.8/4.7) |
| Org set to ZDR / retention below 30 days (Fable 5/5.1, Mythos 5/5.1) | 400 on every request | Fix the org's data-retention configuration - the payload isn't the problem |
-| `tool_choice` `any` / `tool` on Claude Fable 5.1 / Claude Mythos 5.1 / Mythos Preview | 400 | `{type: "auto"}` + name the tool in the prompt (`strict: true` for schema-valid args), or structured outputs |
-| Edited history replayed with thinking blocks (Claude Fable 5.1 / Claude Mythos 5.1, preserved thinking) | 400 `Invalid signature in thinking block ... bound to a different conversation` | Stop editing history - keep the transcript append-only, using mid-conversation `role: "system"` / tool-change messages, turn-scoped `clear_at` reminders that are never deleted, server-side context editing, and summary-only compaction instead of edits; recover once by stripping the named block and every thinking block after it (text and tool calls stay), or `prefix_mismatch_behavior: "drop_block"` |
+| `thinking: {type: "disabled"}` or `budget_tokens` on Claude Opus 5.5 | 400 `"thinking.type.disabled" is not supported for this model` | Omit `thinking`; control depth with `output_config.effort` (default `medium`) |
+| `computer_20251124` tool on Claude Opus 5.5 | 400 `does not support tool types: computer_20251124` | `{type: "computer_toolset_20260801"}` - no beta header, no `name` / display size; update the agent loop for member tool calls |
+| `tool_choice` `any` / `tool` on Claude Fable 5.1 / Claude Mythos 5.1 / Claude Opus 5.5 / Mythos Preview | 400 | `{type: "auto"}` + name the tool in the prompt (`strict: true` for schema-valid args), or structured outputs |
+| Edited history replayed with thinking blocks (Claude Fable 5.1 / Claude Mythos 5.1 / Claude Opus 5.5, preserved thinking) | 400 `Invalid signature in thinking block ... bound to a different conversation` | Stop editing history - keep the transcript append-only, using mid-conversation `role: "system"` / tool-change messages, turn-scoped `clear_at` reminders that are never deleted, server-side context editing, and summary-only compaction instead of edits; recover once by stripping the named block and every thinking block after it (text and tool calls stay), or `prefix_mismatch_behavior: "drop_block"` |
| `thinking.block_binding` without `thinking-binding-controls-2026-08-01` | 400 `block_binding: Extra inputs are not permitted` | Send the beta header where the controls beta is offered (`shared/platform-availability.md`); elsewhere remove `block_binding` and use strip-and-retry |
| `budget_tokens` >= `max_tokens` (older models) | 400 | Ensure `budget_tokens` < `max_tokens` |
| Typo in model ID | 404 | Use valid model ID like `claude-opus-5` |
diff --git a/skills/claude-api/shared/live-sources.md b/skills/claude-api/shared/live-sources.md
index baebdef..6baa4e9 100644
--- a/skills/claude-api/shared/live-sources.md
+++ b/skills/claude-api/shared/live-sources.md
@@ -69,7 +69,7 @@ This file contains WebFetch URLs for fetching current information from platform.
| Topic | URL | Extraction Prompt |
| -------------- | -------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------- |
| Code Execution | `https://platform.claude.com/docs/en/agents-and-tools/tool-use/code-execution-tool.md` | "Extract code execution tool setup, file upload, container reuse, and response handling" |
-| Computer Use | `https://platform.claude.com/docs/en/agents-and-tools/tool-use/computer-use-tool.md` | "Extract computer use tool setup, capabilities, and implementation examples" |
+| Computer Use | `https://platform.claude.com/docs/en/agents-and-tools/tool-use/computer-use-tool.md` | "Extract the computer_toolset_20260801 setup (configs, member tools, batch actions, toolset_name on results), the Compatibility matrix, and the migration steps from computer_20251124" |
| Bash Tool | `https://platform.claude.com/docs/en/agents-and-tools/tool-use/bash-tool.md` | "Extract bash tool schema, reference implementation, and security considerations" |
| Text Editor | `https://platform.claude.com/docs/en/agents-and-tools/tool-use/text-editor-tool.md` | "Extract text editor tool commands, schema, and reference implementation" |
| Memory Tool | `https://platform.claude.com/docs/en/agents-and-tools/tool-use/memory-tool.md` | "Extract memory tool commands, directory structure, and implementation patterns" |
diff --git a/skills/claude-api/shared/model-migration.md b/skills/claude-api/shared/model-migration.md
index def21ea..229c500 100644
--- a/skills/claude-api/shared/model-migration.md
+++ b/skills/claude-api/shared/model-migration.md
@@ -27,6 +27,8 @@ For the latest, authoritative version (with code samples in every supported lang
| Claude Fable 5.1 Migration Checklist | The required vs optional items for Claude Fable 5.1, tagged `[BLOCKS]` / `[TUNE]` |
| Migrating to Claude Fable 5.1 from Claude Fable 5 | Migrating Claude Fable 5 / Claude Opus 5 / Claude Mythos 5 -> Claude Fable 5.1 or Claude Mythos 5.1 (forced `tool_choice` 400s; "preserved thinking" - model-bound blocks and the history-editing check; per-message effort; append-only per-turn reminders; `display: "updates"` progress updates; cheaper cache reads; behavioral re-tuning) |
| Claude Fable 5.1 from Claude Fable 5 Migration Checklist | The required vs optional items for the Claude Fable 5 -> Claude Fable 5.1 move, tagged `[BLOCKS]` / `[TUNE]` |
+| Migrating to Claude Opus 5.5 | Migrating Claude Opus 5 -> Claude Opus 5.5 (thinking can't be disabled; forced `tool_choice` 400s; preserved thinking; computer use via the toolset only; progress updates as thinking blocks; default effort `medium`; broader classifiers; effort tuning + prompting guidance) |
+| Claude Opus 5.5 Migration Checklist | The required vs optional items for Claude Opus 5.5, tagged `[BLOCKS]` / `[TUNE]` |
| Verify the Migration | After edits - runtime spot-check |
| Ground the migration with an eval | User reports a behavioral regression on the new model |
@@ -76,7 +78,7 @@ Not every file that contains the old model ID is a **caller** of the API. Before
| 1 | **Calls the API/SDK** | `client.messages.create(model=...)`, `anthropic.Anthropic()`, request payloads | Swap the model ID **and** apply the breaking-change checklist for the target version (below). |
| 2 | **Defines or serves the model** | Model registries, OpenAPI specs, routing/queue configs, model-policy enums, generated catalogs | The old entry **stays** (the model is still served). Ask whether to (a) add the new model alongside, (b) leave alone, or (c) retire the old model - never blind-replace. **If you can't ask, default to (a): add the new model alongside and flag it** - replacing would de-register a model that's still in production. |
| 3 | **References the ID as an opaque string** | UI fallback constants, capability-gate substring checks, generic test fixtures, label parsers, env defaults | Usually swap the string and verify any parser/regex/substring match handles the new ID - but check the sub-cases below first. |
-| 4 | **Suffixed variant ID** | `claude-<model>-<suffix>` like `-fast`, `-1024k`, `-200k`, `[1m]`, dated snapshots | These are deployment/routing identifiers, not the public model ID. **Do not assume a new-model equivalent exists.** Verify in the registry first; if absent, leave the string alone and flag it. **Exception: `-fast` strings (e.g. `claude-opus-4-6-fast`) are handled by the Fast Mode section below**, which rewrites them to Opus 4.8 plus `speed="fast"` and the `fast-mode-2026-02-01` beta rather than leaving them in place. |
+| 4 | **Suffixed variant ID** | `claude-<model>-<suffix>` like `-fast`, `-1024k`, `-200k`, `[1m]`, dated snapshots | These are deployment/routing identifiers, not the public model ID. **Do not assume a new-model equivalent exists.** Verify in the registry first; if absent, leave the string alone and flag it. **Exception: `-fast` strings (e.g. `claude-opus-4-6-fast`) are handled by the Fast Mode section below**, which rewrites them to Claude Opus 5 plus `speed="fast"` and the `fast-mode-2026-02-01` beta rather than leaving them in place. |
**Bucket 3 sub-cases - before swapping a string reference, check:**
@@ -189,6 +191,7 @@ If you're applying several prompt-tuning edits at once, offer them as a short li
| Claude Mythos Preview (`claude-mythos-preview`) | `claude-mythos-5-1` (Project Glasswing successor) or `claude-fable-5-1` (GA) | Same tokenizer family - mostly a model-ID swap; remove `thinking` config and prefill; see Migrating to Claude Fable 5.1 |
| Claude Fable 5 (`claude-fable-5`) | `claude-fable-5-1` | Same tier, same per-token price, same tokenizer; three breaking changes (forced `tool_choice` 400s, "preserved thinking") - see Migrating to Claude Fable 5.1 from Claude Fable 5 |
| Claude Mythos 5 (`claude-mythos-5`) | `claude-mythos-5-1` | Same path as claude-fable-5 -> claude-fable-5-1; see § Claude Mythos 5.1 under Migrating to Claude Fable 5.1 from Claude Fable 5 |
+| Claude Opus 5 (`claude-opus-5`) | `claude-opus-5-5` - only when the user names Claude Opus 5.5 (launching) | Lower price ($4 / $20 vs $5 / $25), same context window and tokenizer; four breaking changes (thinking can't be disabled, forced `tool_choice` 400s, preserved thinking, computer use via the toolset only) - see Migrating to Claude Opus 5.5 |
| Opus 4.8 | `claude-opus-5` | The current Opus. Two breaking changes (thinking on by default; disabling thinking capped at `high` effort) plus prompt re-tuning - see Migrating to Claude Opus 5 |
| Opus 4.7 | `claude-opus-5` | Apply the Opus 4.8 section (prompt re-tuning, no new breaking changes), then the Claude Opus 5 section |
| Opus 4.6 | `claude-opus-5` | Apply the Opus 4.7 breaking changes, then 4.8 re-tuning, then the Claude Opus 5 section |
@@ -197,7 +200,7 @@ If you're applying several prompt-tuning edits at once, offer them as a short li
| Sonnet 4.0 / 4.5 / 3.7 / 3.5 | `claude-sonnet-5` | Apply the Sonnet 4.6 changes first, then the Claude Sonnet 5 section |
| Haiku 3 / 3.5 | `claude-haiku-4-5` | Fastest and most cost-effective |
-Default to the latest Opus for the caller's tier unless they explicitly chose otherwise. The Opus migrations layer: if you're on Opus 4.6 or older, apply each version's section in order up to your target (e.g. 4.5 -> 4.8 means the 4.6, 4.7, and 4.8 sections in sequence). A 4.7 -> 4.8 move has no new breaking changes - see Migrating to Opus 4.8 below.
+Default to the latest Opus for the caller's tier unless they explicitly chose otherwise - except Claude Opus 5.5, which is launching: migrate to it only when the user names it. The Opus migrations layer: if you're on Opus 4.6 or older, apply each version's section in order up to your target (e.g. 4.5 -> 4.8 means the 4.6, 4.7, and 4.8 sections in sequence). A 4.7 -> 4.8 move has no new breaking changes - see Migrating to Opus 4.8 below.
---
@@ -495,6 +498,7 @@ If the model is now overtriggering a tool or skill, the fix is almost always to
| `claude-opus-4-5` | `claude-opus-5` |
| `claude-opus-4-1` | `claude-opus-5` |
| `claude-opus-4-0` | `claude-opus-5` |
+| `claude-opus-5` | `claude-opus-5-5` - only when the user names Claude Opus 5.5 |
| `claude-mythos-preview` | `claude-mythos-5-1` (Project Glasswing) or `claude-fable-5-1` |
| `claude-fable-5` | `claude-fable-5-1` |
| `claude-mythos-5` | `claude-mythos-5-1` |
@@ -512,6 +516,7 @@ If the code uses the `AnthropicBedrockMantle` client (Python `anthropic[bedrock]
|---|---|
| `claude-opus-4-8` | `anthropic.claude-opus-4-8` |
| `claude-opus-5` | `anthropic.claude-opus-5` |
+| `claude-opus-5-5` | `anthropic.claude-opus-5-5` |
| `claude-fable-5-1` | `anthropic.claude-fable-5-1` |
| `claude-fable-5` | `anthropic.claude-fable-5` |
| `claude-mythos-5-1` | `anthropic.claude-mythos-5-1` (us-east-1 only, not publicly listed) |
@@ -700,11 +705,11 @@ Beyond resolution, Opus 4.7 also improves on low-level perception (pointing, mea
Requests that involve prohibited or high-risk topics may lead to refusals.
-### Fast Mode: Claude Opus 5 / Opus 4.8 only
+### Fast Mode: Claude Opus 5 / Claude Opus 5.5 / Opus 4.8 only
-Fast mode is available on Claude Opus 5 and Opus 4.8. Only surface this if the caller's code actually uses fast mode (e.g. `model="claude-opus-4-6-fast"`, or `speed="fast"` on an unsupported model); if the word "fast" does not appear in the code, say nothing about Fast Mode.
+Fast mode is available on Claude Opus 5, Claude Opus 5.5, and Opus 4.8. Only surface this if the caller's code actually uses fast mode (e.g. `model="claude-opus-4-6-fast"`, or `speed="fast"` on an unsupported model); if the word "fast" does not appear in the code, say nothing about Fast Mode.
-When you see `model="claude-opus-4-6-fast"` (or any retired `-fast` model string), **the migration edit is** to move the fast-mode traffic onto Claude Opus 5, the current fast-capable default (Opus 4.8 also works if the caller is staying on that tier):
+When you see `model="claude-opus-4-6-fast"` (or any retired `-fast` model string), **the migration edit is** to move the fast-mode traffic onto Claude Opus 5, the current fast-capable default (Opus 4.8 also works if the caller is staying on that tier; Claude Opus 5.5 supports it too, at $8 / $40 - its fast-mode docs flip after the model launch, confirm before quoting - use it when the user names it):
```python
# Request fast mode on Claude Opus 5.
@@ -715,7 +720,7 @@ client.beta.messages.create(
)
```
-That is: switch the model to Claude Opus 5 (or Opus 4.8) and request fast mode the supported way, using the beta `client.beta.messages....` endpoint, the `fast-mode-2026-02-01` beta flag, and `speed="fast"` as a top-level request parameter (per-language form in SKILL.md § Fast Mode). Opus 4.7 fast mode has also been removed, so do not land on Opus 4.7 either. Do **not** leave the code on a retired `-fast` model string - the failure mode differs by version: `claude-opus-4-6-fast` is retired and the API **silently falls back** to standard Opus 4.6 (no error - the caller loses fast-mode speed without noticing); `claude-opus-4-7-fast` and `speed="fast"` on Opus 4.7 instead return an **API error** (hard failure - requests break outright rather than degrading). Either way, migrate to Opus 4.8 fast mode now.
+That is: switch the model to Claude Opus 5 (or Opus 4.8) and request fast mode the supported way, using the beta `client.beta.messages....` endpoint, the `fast-mode-2026-02-01` beta flag, and `speed="fast"` as a top-level request parameter (per-language form in SKILL.md § Fast Mode). Opus 4.7 fast mode has also been removed, so do not land on Opus 4.7 either. Do **not** leave the code on a retired `-fast` model string - the failure mode differs by version: `claude-opus-4-6-fast` is retired and the API **silently falls back** to standard Opus 4.6 (no error - the caller loses fast-mode speed without noticing); `claude-opus-4-7-fast` and `speed="fast"` on Opus 4.7 instead return an **API error** (hard failure - requests break outright rather than degrading). Either way, migrate to a supported fast-mode model (Claude Opus 5 by default) now.
### Behavioral shifts (prompt-tunable)
@@ -1599,7 +1604,7 @@ Keep passing thinking blocks back unchanged when you switch models - the API dro
The published docs file this and breaking change 2 together under *preserved thinking* ("pass blocks back unchanged and let the API decide which the model can use"); this one is the conversation check - editing earlier turns invalidates every later thinking block. The API field names for it say `prefix_mismatch_behavior` / `prefix_binding_mismatch` - the same check.
-A Claude Fable 5.1 thinking block's `signature` also records the conversation prefix that produced it - the top-level `system` prompt, the set of tools in `tools`, and every message before the block (with server-side compaction, the prefix starts at the most recent compaction block) - plus a chain to the previous thinking block across turns (earlier thinking blocks aren't part of the prefix, but each block records the one before it, which is why blocks can be removed from the *front* of the history and not from the middle). When the transcript comes back, the API checks that this prefix is unchanged. Claude Code, claude.ai, Managed Agents, and the Agent SDK keep the prefix intact for you; **if your code builds the `messages` array itself, check it before migrating** (the three-step check is below). **Who is enforced:** new accounts **created on or after August 31, 2026** (Claude API organizations, Amazon Bedrock accounts, Google Cloud projects, Microsoft Foundry resources). Anthropic plans to enforce it for every account on future models, so adopt the patterns now even if your account isn't enforced today. For accounts created earlier the API *records* the mismatch but acts on it only when the request opts in: setting `thinking.block_binding.prefix_mismatch_behavior` - **any value, including `"error"`, opts the request into enforcement**, which is also how you test from an older organization - or sending the `thinking-binding-controls-2026-08-01` header alone, which opts the request into the beta's default, `drop_block`. If you ship a tool or framework that people run with their own API key, test with the field set: your users on new organizations are enforced before you are. To see whether your own organization is enforced by default, send a request that edits history without the beta header - a 400 that names the header means it is. Platform note: the opt-in controls themselves (the beta header, `prefix_mismatch_behavior`, `input_transformations`) are on the Claude API and Claude Platform on AWS at launch, arrive per model on Amazon Bedrock and Google Cloud (until then the header is rejected there), and aren't offered on Microsoft Foundry - on a platform without the controls the opt-in test path doesn't apply and recovery is strip-and-retry (`shared/platform-availability.md` has the matrix).
+A Claude Fable 5.1 thinking block's `signature` also records the conversation prefix that produced it - the top-level `system` prompt, the set of tools in `tools`, and every message before the block (with server-side compaction, the prefix starts at the most recent compaction block) - plus a chain to the previous thinking block across turns (earlier thinking blocks aren't part of the prefix, but each block records the one before it, which is why blocks can be removed from the *front* of the history and not from the middle). When the transcript comes back, the API checks that this prefix is unchanged. Claude Code, claude.ai, Managed Agents, and the Agent SDK keep the prefix intact for you; **if your code builds the `messages` array itself, check it before migrating** (the three-step check is below). **Who is enforced:** new accounts **created on or after August 31, 2026** (Claude API organizations, Amazon Bedrock accounts, Google Cloud projects, Microsoft Foundry resources). Enforcement scope is decided per model - Claude Opus 5.5 also enforces it for new accounts only - so make your application compatible regardless of your account's age: the same patterns keep the prompt cache warm, and you can test against the check from any account by sending `prefix_mismatch_behavior`. For accounts created earlier the API *records* the mismatch but acts on it only when the request opts in: setting `thinking.block_binding.prefix_mismatch_behavior` - **any value, including `"error"`, opts the request into enforcement**, which is also how you test from an older organization - or sending the `thinking-binding-controls-2026-08-01` header alone, which opts the request into the beta's default, `drop_block`. If you ship a tool or framework that people run with their own API key, test with the field set: your users on new organizations are enforced before you are. To see whether your own organization is enforced by default, send a request that edits history without the beta header - a 400 that names the header means it is. Platform note: the opt-in controls themselves (the beta header, `prefix_mismatch_behavior`, `input_transformations`) are on the Claude API and Claude Platform on AWS at launch, arrive per model on Amazon Bedrock and Google Cloud (until then the header is rejected there), and aren't offered on Microsoft Foundry - on a platform without the controls the opt-in test path doesn't apply and recovery is strip-and-retry (`shared/platform-availability.md` has the matrix).
**What invalidates every later thinking block:**
@@ -1841,7 +1846,7 @@ The second tells it to hold the scope the user set:
- [ ] **[BLOCKS]** Coming from an Opus-tier or older model (not from Claude Fable 5): apply the Claude Fable 5.1 Migration Checklist above (the Opus-tier -> Fable migration) first, plus § Coming from Claude Opus 5 - `thinking: {type: "disabled"}` now 400s at any effort, between-tool narration moves into `thinking` blocks, ZDR is lost, price doubles
- [ ] **[BLOCKS]** Data retention: 30-day retention required (Covered Model; ZDR only if expressly authorized by Anthropic) - a ZDR org gets `400 invalid_request_error` on every request, as on Claude Fable 5; check the retention configuration before debugging the payload
- [ ] **[BLOCKS]** Keep passing `thinking` blocks back unchanged on every turn, including empty ones and `redacted_thinking` - the history-editing check rejects edited history
-- [ ] **[BLOCKS]** Preserved thinking / the history-editing check (new accounts created on/after 2026-08-31 on every platform, and any request that sets `prefix_mismatch_behavior` or sends the controls beta header; later models enforce it for everyone): stop editing history between requests - freeze the top-level `system`, use `role: "system"` messages for mid-session instructions, `tool_addition`/`tool_removal` for tool changes, turn-scoped (`clear_at`) system messages - or, without that beta, retained user-message text blocks - appended after the tool results and never deleted, for per-turn reminders, server-side context editing / compaction (summary-only if client-side) for trimming, `file_id` for cross-turn files. Run the three-step check on a platform offering the controls beta (`shared/platform-availability.md`) (`prefix_mismatch_behavior: "drop_block"` + log `input_transformations`; fix every `prefix_binding_mismatch`, `model_binding_mismatch` after a model switch is expected; `"error"` in CI), then pick a production setting and monitor it. If you ship a tool others run with their own key, test with the field set. Keep-tail and background compaction need `"drop_block"` (per request - keep sending it) or stripped thinking on the retained turns; never compact mid tool round
+- [ ] **[BLOCKS]** Preserved thinking / the history-editing check (new accounts created on/after 2026-08-31 on every platform, and any request that sets `prefix_mismatch_behavior` or sends the controls beta header; enforcement scope is decided per model, and Claude Opus 5.5 also enforces it for new accounts only): stop editing history between requests - freeze the top-level `system`, use `role: "system"` messages for mid-session instructions, `tool_addition`/`tool_removal` for tool changes, turn-scoped (`clear_at`) system messages - or, without that beta, retained user-message text blocks - appended after the tool results and never deleted, for per-turn reminders, server-side context editing / compaction (summary-only if client-side) for trimming, `file_id` for cross-turn files. Run the three-step check on a platform offering the controls beta (`shared/platform-availability.md`) (`prefix_mismatch_behavior: "drop_block"` + log `input_transformations`; fix every `prefix_binding_mismatch`, `model_binding_mismatch` after a model switch is expected; `"error"` in CI), then pick a production setting and monitor it. If you ship a tool others run with their own key, test with the field set. Keep-tail and background compaction need `"drop_block"` (per request - keep sending it) or stripped thinking on the retained turns; never compact mid tool round
- [ ] **[TUNE]** Fallbacks: keep server-side `fallbacks` (targets `claude-opus-4-8` / `claude-opus-5`; routing unpublished) or the SDK middleware; the fallback model can't read 5.1 thinking blocks (dropped, unbilled); fallback credit works as on Claude Fable 5
- [ ] **[TUNE]** Adopt `thinking: {type: "adaptive", display: "updates"}` with `thinking-display-updates-2026-08-18` (all platforms) if users watch long tool-calling turns; render non-empty `thinking` blocks as status lines, handle the interrupted-response sentinel, echo them back unchanged
- [ ] **[TUNE]** Adopt per-message effort (`mid-conversation-output-config-2026-07-01`; also on Claude Opus 5) where a loop mixes hard and routine steps - lowering is reliable, raising wants a big jump; re-run the effort sweep (`high` default; `medium` as cost control; `xhigh`/`max` only for capability-sensitive work; `low` often beats below-frontier models on cost per task); size `max_tokens` for `high`+
@@ -1852,9 +1857,215 @@ The second tells it to hold the scope the user set:
---
+## Migrating to Claude Opus 5.5
+
+> **Model ID `claude-opus-5-5` is authoritative as written here.** When the user asks to migrate to Claude Opus 5.5, write `model="claude-opus-5-5"` exactly. Do **not** WebFetch to verify - this guide is the source of truth for migration target IDs. The corresponding entry exists in `shared/models.md`.
+
+Claude Opus 5.5 succeeds Claude Opus 5 in the Opus line for long-running agentic coding and knowledge work, **at a lower price** - $4 / $20 per MTok input / output against Claude Opus 5's $5 / $25. Same 1M token context window (default and maximum), same 128K max output, same tokenizer as Claude Opus 5 (token counts unchanged; coming from a pre-Opus-4.7 model, follow the tokenizer guidance in the Claude Opus 5 section). Knowledge cutoff June 2026. Available at launch on the Claude API (`claude-opus-5-5`), Amazon Bedrock (`anthropic.claude-opus-5-5`), Claude Platform on AWS, Google Cloud, and Microsoft Foundry (all as `claude-opus-5-5`; on Foundry both the Anthropic-hosted and Microsoft-hosted paths, the latter subject to its own launch-day go/no-go); Claude Opus 5 stays available on all of them. Existing Claude Opus 5 prompts should perform well out of the box; the Claude Opus 5 prompting patterns below remain a reasonable starting point.
+
+**Migrate to Claude Opus 5.5 when the user names it.** This section is written ahead of the launch: the skill's mandated default (`claude-opus-5`) and the pricing table in SKILL.md don't move until then. It is layered on top of the Claude Opus 5 migration above - a caller coming from Opus 4.8 or older applies § Migrating to Claude Opus 5 first (Opus 4.7 or older: the sections before that), with two exceptions to what that section (and the earlier ones) say: Claude Opus 5's "thinking can be disabled at `high` or below" does not carry over, and neither does the acceptance of the earlier `computer_20251124` tool (breaking change 4 below). Coming from Claude Sonnet 5: the request surface already matches (adaptive thinking, no sampling parameters, no prefill) - apply this section on top of the Claude Sonnet 5 code, re-baselining for Opus-tier pricing and rate limits.
+
+**What changes, in one line:** four breaking changes for code running on Claude Opus 5 (thinking can't be disabled; forced `tool_choice` 400s; thinking blocks are tied to the model and the conversation - "preserved thinking"; the `computer_20251124` tool 400s - use the computer toolset), one response-shape change that fails no request (text between tool calls comes back in `thinking` blocks), a **default effort of `medium`** where Claude Opus 5's is `high`, and a broader safety-classifier set (`bio` and `reasoning_extraction` join `cyber`). The first three breaking changes are the same mechanisms Claude Fable 5.1 introduced - the sections below give the Claude Opus 5.5 specifics and point at § Migrating to Claude Fable 5.1 from Claude Fable 5 for the shared mechanics rather than repeating them. Everything else in the Claude Opus 5 request surface carries over: mid-conversation system messages and per-message effort (which some of the tips below use), mid-conversation tool changes, task budgets, compaction, the 512-token minimum cacheable prompt, batch, the Files API, PDF support, vision, and the server-side and client-side tools.
+
+### Breaking change 1: thinking can't be disabled
+
+On Claude Opus 5, thinking is on by default and `thinking: {type: "disabled"}` is accepted at effort `high` or below. On Claude Opus 5.5 thinking is **always on**: `{"type": "disabled"}` and `{"type": "enabled", "budget_tokens": N}` both return a 400 `invalid_request_error` at every effort level, with no beta header involved:
+
+```text
+"thinking.type.disabled" is not supported for this model. Use "thinking.type.adaptive" and "output_config.effort" to control thinking behavior.
+```
+
+(`"thinking.type.enabled" is not supported for this model. ...` for the budget form.) Omit the `thinking` field or send `{type: "adaptive"}`, which is equivalent. **Effort is now the control for how much the model thinks, and therefore for latency and cost** (§ Choosing an effort level below). Migrate a route that disables thinking or sets a budget as follows:
+
+1. **Remove the `thinking` field** (or set it to `{type: "adaptive"}`).
+2. **If time to first token matters, set `output_config.effort` to `low`.** At `low` the model keeps its thinking short; how often it skips thinking altogether depends on the prompts. Measure, and move to `medium` if quality drops. A system-prompt line such as *"Answer directly without deliberating."* can reduce thinking further (and with it TTFT and cost) - measure quality on your own use case before keeping it, since less thinking can cost accuracy.
+3. **Size `max_tokens` for the thinking as well as the reply.** Thinking counts toward `max_tokens` even though its text isn't returned under the default `display`, so a limit sized for a no-thinking route cuts replies off. For long agentic coding turns, 64K has worked well.
+4. **Read the response by block `type`, not position.** A response can begin with one or more `thinking` blocks; under the default `display: "omitted"` they come back with an empty `thinking` string. Set `display: "summarized"` for a readable summary of the reasoning. Pass `thinking` blocks back unmodified in tool-use loops.
+
+```python
+# Before - accepted on Claude Opus 5, 400 on Claude Opus 5.5
+client.messages.create(
+ model="claude-opus-5",
+ max_tokens=16000,
+ thinking={"type": "disabled"},
+ messages=[{"role": "user", "content": "..."}],
+)
+
+# After - thinking is always on; effort is the control
+client.messages.create(
+ model="claude-opus-5-5",
+ max_tokens=16000,
+ output_config={"effort": "low"},
+ messages=[{"role": "user", "content": "..."}],
+)
+```
+
+**Prompts written for thinking disabled.** Three follow-ups if the Claude Opus 5 integration ran with thinking off: (a) start at `low` and measure, as above; (b) **remove instructions that stood in for thinking** - a prompt that asked the model to write its reasoning into the response text as a substitute for thinking should go, and the reasoning read from `display: "summarized"` blocks instead; a prompt that pushes the model to reproduce its internal reasoning in the response can be **declined** with `stop_details.category: "reasoning_extraction"`; (c) re-test the two thinking-disabled mitigations from § Two failure modes when thinking is disabled under Claude Opus 5 - both address artifacts that appeared only with thinking off, so check whether the combined "brief sentence before a tool call / say so if no tool fits / no internal XML tags" instruction is still needed, and **delete any rule telling the model not to think either way** (it can't comply, and such rules increase tag leakage).
+
+### Breaking change 2: forced tool use is rejected
+
+As on Claude Fable 5.1: `tool_choice: {"type": "any"}` and `{"type": "tool", "name": "..."}` return a 400 `invalid_request_error` (`tool_choice: type "tool" and "any" are not supported for this model.`) on the Messages API, the Message Batches API, and the token-counting endpoint, where Claude Opus 5 accepts both. `{"type": "auto"}` (the default) and `{"type": "none"}` are unchanged; `disable_parallel_tool_use: true` still works with `auto` but now means *at most* one call. Migrate by intent - the full patterns (steering from the prompt, `strict: true` for schema-valid arguments, structured outputs for extraction, the advisor tool) are under § Breaking change 1: forced tool use is rejected in § Migrating to Claude Fable 5.1 from Claude Fable 5. The two most common:
+
+- **Steering toward a tool:** `tool_choice: {"type": "auto"}` plus the expectation in the prompt ("Use the `get_weather` tool to answer"), with `strict: true` on the tool definition (schema sets `additionalProperties: false`) so the arguments match the schema. Because `auto` does not guarantee a call, **check that one was made and retry if it wasn't.**
+- **Extracting structured data:** if the forced call existed only to get JSON back, replace it with structured outputs (`output_config.format`).
+
+```python
+# Before - 400 on Claude Opus 5.5
+response = client.messages.create(
+ model="claude-opus-5",
+ max_tokens=1024,
+ tools=tools,
+ tool_choice={"type": "tool", "name": "get_weather"},
+ messages=[{"role": "user", "content": "What's the weather in Paris?"}],
+)
+
+# After - auto + strict tool use, steering in the prompt, and a check that the call happened
+response = client.messages.create(
+ model="claude-opus-5-5",
+ max_tokens=1024,
+ tools=[{**tool, "strict": True} for tool in tools],
+ tool_choice={"type": "auto"},
+ messages=[{"role": "user", "content": "What's the weather in Paris? Use the get_weather tool."}],
+)
+if not any(block.type == "tool_use" for block in response.content):
+ ... # retry, or fall back to a text answer
+```
+
+### Breaking change 3: thinking blocks are tied to the model and the conversation
+
+Both halves of "preserved thinking" from Claude Fable 5.1 apply to Claude Opus 5.5; the mechanics (what invalidates a block, the `drop_block` request shape, `input_transformations`, the three-step audit, the append-only replacements table, which client-side compaction shapes break) are under § Breaking change 2 and § Breaking change 3 in § Migrating to Claude Fable 5.1 from Claude Fable 5 and apply verbatim. What is specific to Claude Opus 5.5:
+
+- **Model binding - who reads whose blocks.** Claude Opus 5.5 reads thinking blocks from Claude Opus 5 and earlier Opus, Sonnet, and Haiku models - a conversation that *moves onto* `claude-opus-5-5` keeps its reasoning - but **not** from any Fable or Mythos model. In the other direction, on the Claude API only Claude Fable 5.1 and Claude Mythos 5.1 read a Claude Opus 5.5 block; **no other model does** - so a router switch, a client-side retry on another model, or a classifier-refusal fallback (server-side or SDK middleware) to Claude Opus 5 / Claude Opus 4.8 runs the turns after the switch without Claude Opus 5.5's reasoning. The API drops what the target can't read before the model sees it: the request succeeds, dropped blocks aren't billed, and with the `thinking-binding-controls-2026-08-01` header the drop is reported in `input_transformations` with `reason: "model_binding_mismatch"`. Whether Claude Fable 5.1 / Claude Mythos 5.1 also keep Claude Opus 5.5's blocks on Amazon Bedrock and Google Cloud is open at launch - the docs claim it for the Claude API only. Keep passing blocks back unchanged when you switch models; don't strip them yourself.
+- **Conversation binding - who is enforced.** Same posture as Claude Fable 5.1: on every platform the prefix check (the `system` prompt, the `tools` array, and every earlier message must be byte-identical to when the block was produced) is enforced by default for accounts **created on or after August 31, 2026, 00:00 UTC** - a replayed block after such an edit is a 400. Older accounts opt in by setting `thinking.block_binding.prefix_mismatch_behavior` (`"error"` or `"drop_block"`, beta `thinking-binding-controls-2026-08-01`). Claude Code, claude.ai, Managed Agents, and the Agent SDK keep the prefix intact; **if your code builds `messages` itself, run the three-step check before migrating** - and do it now even on an exempt account, because it also raises prompt-cache hit rates. The three edits that break the prefix and their append-only replacements: a per-turn reminder injected and later deleted, or a system prompt changed mid-session (append a mid-conversation `role: "system"` message instead; for a one-turn reminder the `clear_at: "next_user_message"` form under beta `mid-conversation-system-clear-at-2026-08-21` is a limited beta - without it, append the reminder as a text block after the `tool_result` blocks and leave earlier copies in place); tools added or removed mid-session (declare the full set at session start and send `tool_addition` / `tool_removal` blocks, beta `mid-conversation-tool-changes-2026-07-01`); and compaction that summarizes older turns while replaying newer ones with their thinking blocks (use server-side compaction or context editing - the on-demand `compaction` parameter under beta `compact-2026-09-04`, offered on the Claude API, Claude Platform on AWS, Google Cloud, and Microsoft Foundry but not yet Amazon Bedrock, is designed to keep the retained turns' blocks valid after the swap - or client-side *simple* compaction that replaces the whole history with a summary and replays no earlier thinking, or set `drop_block`). Two compaction details that follow Claude Fable 5.1: a threshold-compaction request with custom `instructions` summarizes from the visible conversation only - earlier thinking blocks are not part of the summarizer's input, so tell it what the summary must retain (on-demand compaction's summarizer reads earlier thinking with or without `instructions`); and any assistant turn you re-insert after a compaction block needs its `thinking` / `redacted_thinking` blocks removed, or `drop_block` set.
+
+```http
+POST /v1/messages
+anthropic-beta: thinking-binding-controls-2026-08-01
+
+{"model": "claude-opus-5-5", "max_tokens": 64000,
+ "thinking": {"type": "adaptive", "block_binding": {"prefix_mismatch_behavior": "drop_block"}},
+ "messages": [ ...full history with thinking blocks replayed verbatim... ]}
+```
+
+### Breaking change 4: computer use only through the computer toolset
+
+> **Confirm at launch.** The EAP guide marks the accepted computer-use tool versions and the toolset's platform availability as not final; the launch docs state the rule below. Re-check `shared/live-sources.md` -> Computer use before promising it on a partner platform.
+
+Claude Opus 5 accepts computer use both as the `computer_toolset_20260801` toolset and, with the `computer-use-2025-11-24` beta header, as the earlier `computer_20251124` tool. **Claude Opus 5.5 accepts only the toolset**: a `tools` entry of type `computer_20251124` returns a 400 `invalid_request_error` that names the rejected type and then lists the accepted ones after `Did you mean one of` (it begins `'claude-opus-5-5' does not support tool types: computer_20251124.`). The toolset is GA on the Claude API and Google Cloud with no beta header; other platforms offer only the earlier beta versions today, so an integration on those platforms has no Claude Opus 5.5 computer-use path until they add the toolset - check the computer use tool's Compatibility section (`shared/tool-use-concepts.md` § Computer Use has the toolset summary). This is more than a `tools`-entry swap, so make and test the change on Claude Opus 5 first (it accepts both forms):
+
+- **Request:** drop the beta header and the beta client namespace; the entry is `{"type": "computer_toolset_20260801"}` with **no `name`** and no `display_width_px` / `display_height_px`; an optional `configs` map turns individual member tools on or off (`{"zoom": {"enabled": false}}`). All 17 members, `zoom` included, are on by default. The entry can't share a request with a `computer_20251124` entry or another tool named `computer`.
+- **Agent loop:** Claude's calls are `tool_use` blocks whose `name` is the member (`screenshot`, `left_click`, `type`, `zoom`, ...) - **the action is the block's `name`, not `input.action`** - carrying `"toolset_name": "computer"`, and there can be **several per turn** (a batch action), each its own block. Return one `tool_result` per `tool_use`, matched by `tool_use_id`, all in the next `user` message, **every one echoing `"toolset_name": "computer"`** (a result that omits it is rejected); only `screenshot` and `zoom` results need an image, a short `OK` is enough for the rest. Coordinates are in the pixel space of the full screenshots you return, also after a `zoom`. Screenshots must already fit the model's image limits (the toolset takes no display dimensions and the API doesn't downscale for you).
+
+```python
+# Before - 400 on Claude Opus 5.5
+client.beta.messages.create(
+ model="claude-opus-5",
+ max_tokens=4096,
+ betas=["computer-use-2025-11-24"],
+ tools=[{"type": "computer_20251124", "name": "computer",
+ "display_width_px": 1024, "display_height_px": 768}],
+ messages=[{"role": "user", "content": "Open the display settings."}],
+)
+
+# After - no beta header; the toolset entry takes no name or display size
+client.messages.create(
+ model="claude-opus-5-5",
+ max_tokens=4096,
+ tools=[{"type": "computer_toolset_20260801"}],
+ messages=[{"role": "user", "content": "Open the display settings."}],
+)
+```
+
+Integrations already on the toolset, and the browser use toolset (`browser_toolset_20260801`), need no change.
+
+### Text between tool calls comes back in thinking blocks
+
+On Claude Opus 5, the short notes the model writes between tool calls (what it just found, what it's doing next) come back as `text` blocks. On Claude Opus 5.5, as on Claude Fable 5.1, notes longer than a sentence or two come back as **progress-update `thinking` blocks**, at most one before each tool call, and under the default `display: "omitted"` their text is empty - no request fails, but a client that renders only `text` blocks goes quiet for the length of a long agentic turn. Fix: set `thinking.display: "updates"` (beta `thinking-display-updates-2026-08-18`) to get a short summary of each note as text while reasoning stays hidden (`"summarized"` returns both, mixed), render each non-empty `thinking` block ahead of the `tool_use` it precedes, and pass the blocks back unchanged - the consumption rules (streaming `thinking_delta`, the interrupted-work sentinel, zero-or-more per response) are under addition 3 in § New API features of § Migrating to Claude Fable 5.1 from Claude Fable 5:
+
+```http
+POST /v1/messages
+anthropic-beta: thinking-display-updates-2026-08-18
+
+{"model": "claude-opus-5-5", "max_tokens": 64000,
+ "thinking": {"type": "adaptive", "display": "updates"},
+ "tools": [...],
+ "messages": [{"role": "user", "content": "Review the PRs open against our billing service."}]}
+```
+
+Three levers on what users see:
+
+1. **Receive them** - `display: "updates"` as above (a `"summarized"` display returns them too, mixed with the reasoning summaries).
+2. **If the model may need to hand the user something *verbatim*** partway through a long turn - a code snippet, an exact value - give it a simple tool for sending the user a message and tell it to reserve the tool for that content. Declare the tool in `tools` from the **first** request of the session: adding it later edits the conversation's prefix and invalidates earlier thinking blocks (breaking change 3).
+3. **For more frequent or predictable updates** - a one-line statement of intent before the first tool call and a short recap at the end - say so in the system prompt: when you want user-facing text and what it should contain. The model follows such instructions reasonably well; this helps most in pair programming and other human-in-the-loop work.
+
+### Choosing an effort level - the default is `medium`, and the levels don't map 1:1 from Claude Opus 5
+
+Effort is the main control for how much Claude Opus 5.5 thinks, and with adaptive-only thinking it is the first setting to adjust when trading off intelligence, latency, and cost. Two things change from Claude Opus 5:
+
+- **The API default is `medium`** (Claude Opus 5 and earlier Opus models default to `high`), so a request that omits `effort` now runs one level lower than it did. **Set `effort` explicitly** and re-run the sweep rather than carrying the Claude Opus 5 setting over. Effort names don't mean the same amount of thinking across models: in Anthropic's testing, Claude Opus 5.5 at `medium` exceeds Claude Opus 5 at `high` on coding and knowledge-work evaluations, and on several coding evaluations `low` comes close to it at much lower cost. Start at `medium` and test the neighboring levels; reserve `xhigh` and `max` for work where you have measured a quality gain (all five levels are supported; `max` is uncapped).
+- **At a given level, Claude Opus 5.5 tends to think more per turn than Claude Opus 5**, especially at `xhigh` and `max`. If you keep the `effort` value you set for Claude Opus 5, expect longer turns and more output tokens. To get less thinking, **lower the effort level before adding "think less" instructions** - lowering effort reduces thinking, cost, and latency more reliably than prompting does. Set `max_tokens` with room for the thinking as well as the reply (thinking counts toward it even though the text isn't returned - a limit sized for Claude Opus 5 with thinking off can cut replies off; 64K is a reasonable starting point for long agentic coding turns).
+
+Change effort for individual turns without invalidating the prompt cache with a **per-message effort change** (beta `mid-conversation-output-config-2026-07-01`; the request shape is addition 1 under § New API features of § Migrating to Claude Fable 5.1 from Claude Fable 5, and Claude Opus 5 already supports it). Changing the **top-level** `effort` between requests does invalidate the prompt cache.
+
+### Safeguards - a broader classifier set than Claude Opus 5
+
+Claude Opus 5.5 runs cybersecurity **and biology** safety classifiers similar to Claude Fable 5.1's; coming from Claude Opus 5, the biology classifier is new. Everyday health and educational questions are unaffected, but requests the classifier treats as dual-use biology research (virology, toxicology, molecular design) are declined; on the cybersecurity side, finding vulnerabilities in source code is allowed. Separately - also new relative to Claude Opus 5 - a request that tries to get the model to reproduce its internal reasoning in the response text can be declined with `stop_details.category: "reasoning_extraction"`; if a prompt does this (for example, to get visible reasoning with thinking off), remove the instruction, set `display: "summarized"`, and read the `thinking` blocks. **`reasoning_extraction` declines are not retried on a fallback model.**
+
+A classifier decline arrives as a normal HTTP 200 with `stop_reason: "refusal"` and a `stop_details` object naming the category (`"cyber"`, `"bio"`, `"reasoning_extraction"`, ...; branch on `stop_reason`, treat `stop_details` as informational - the full handling is § `refusal` stop reason under § Migrating to Claude Fable 5.1). A refusal before any output is not billed, but the request still counts against your rate limits (the billing wording is under revision at launch - confirm). Retry on another model with server-side fallbacks - `fallbacks: "default"` under beta `server-side-fallback-2026-07-01` retries on the model Anthropic recommends for that category, the array form under `server-side-fallback-2026-06-01` names your own targets (§ New API features under § Migrating to Claude Opus 5 has both shapes; the permitted targets for Claude Opus 5.5 are open at launch - expect Claude Opus 5 / claude-opus-4-8), the SDK middleware on platforms without server-side fallback, or your own retry. A fallback model runs without Claude Opus 5.5's thinking blocks (breaking change 3). **Ship the opt-in from day one**, as the Claude Fable 5.1 section says.
+
+The classifiers can still flag benign requests - the fallback opt-in is what keeps a false positive from becoming an outage. (The EAP guide's prompt-side workarounds for specific false positives were not carried into the launch docs - don't cite them.)
+
+### What carries over unchanged from Claude Opus 5 - and what's open at launch
+
+- **Feature set:** per-message effort (beta), mid-conversation system messages (no header) and tool changes (beta), task budgets, compaction (including the on-demand `compaction` parameter, beta `compact-2026-09-04`, which has its own docs), prompt caching with the 512-token minimum, batch processing (up to 300K output tokens with the `output-300k-2026-03-24` beta), the Files API, PDF support, vision, structured outputs, strict tool use, and the same server-side and client-side tools - except computer use, which needs the toolset (breaking change 4). Programmatic tool calling lists the model.
+- **Pricing:** $4 / $20 per MTok; 5-minute cache writes $5 and 1-hour cache writes $8 (derived from the standard 1.25x / 2x multipliers - confirm at launch); **cache reads $0.20 per MTok (0.05x base input)**; batch $2 / $10. The cache-read discount is deeper than Claude Opus 5's, so long agentic sessions that re-read a cached prefix save more, and a miss costs relatively more - keeping the cache warm (per-message effort, append-only histories, the keep-alive patterns in `shared/prompt-caching.md`) matters more.
+- **Rate limits:** a separate pool from Claude Opus 5's (own per-tier numbers - may instead share Claude Opus 5's group; confirm at launch). Re-check your tier's Claude Opus 5.5 limits before moving volume.
+- **Priority Tier:** not supported (as Claude Opus 5).
+- **Fast mode:** research preview on the Claude API only (not Bedrock, Claude Platform on AWS, Google Cloud, or Foundry), `speed: "fast"` under beta `fast-mode-2026-02-01`, at **$8 / $40 per MTok** (2x the standard price, the same multiple as Claude Opus 5's $10 / $50) - the fast-mode docs flip after the model's own launch, so confirm before quoting.
+- **Data retention / ZDR:** the launch docs say nothing new - treat Claude Opus 5.5 as Claude Opus 5 here, and check `shared/platform-availability.md` at launch.
+- **SDK constants:** `Model.ClaudeOpus5_5` (C#), `anthropic.ModelClaudeOpus5_5` (Go), `Model.CLAUDE_OPUS_5_5` (Java, PHP), `Anthropic::Model::CLAUDE_OPUS_5_5` (Ruby) - published with each SDK's launch release; the bare string `"claude-opus-5-5"` works everywhere before then.
+
+### Capability improvements versus Claude Opus 5
+
+**Cheaper per solved task, not just per token.** On many coding, analysis, and vision tasks, Claude Opus 5.5 at its default effort matched or beat Claude Opus 5 while using fewer tokens, and its price per token is 20% lower than Claude Opus 5's (60% lower for cache reads) - so expect the cost per solved task to be significantly lower for most tasks, and re-baseline `shared/cost-optimization.md`'s Claude Opus 5 figures rather than scaling them by the list price alone.
+
+- **Agentic coding and code review:** the largest measured gains are on multistep work in a real codebase (carrying a change through a large repository until its tests pass) - in Anthropic's testing, at its default `medium` effort it matched or beat Claude Opus 5's `high`-effort results on such tasks, in fewer steps and with about half the tokens - and on code review, where it catches more bugs with fewer false alarms. It explains its changes in plain language, which makes its work easier to review and trust. It tends to finish the same task with fewer tokens.
+- **Knowledge work:** a more reliable analyst - much less likely than Claude Opus 5 to state a figure or cite a source the inputs don't support (citations pointed at the right source much more often in one customer's measurement); at `medium` it produced better long analytical deliverables than Claude Opus 5 at `high` with roughly 40% fewer output tokens; better at building and auditing financial models; more detail-oriented on large inputs (a date in a long thread that falls on the wrong weekday, a chart in a deck that doesn't match the figures) **without more false-positive flags**.
+- **Communication and writing:** clearer prose, most noticeably in how it reports on agentic work - its updates while it works and its summary when it finishes say plainly what it did, what it found, and what it needs from you, with less jargon and fewer stock phrases.
+- **Charts, diagrams, screenshots, and computer use:** reads visual material more accurately at every effort level without extra tooling - in Anthropic's testing, even at `low` it read charts more accurately than Claude Opus 5 at its highest effort, at roughly a tenth of the output tokens per chart (Claude Opus 5 read charts well only by running code to crop, zoom, and measure) - values on dense charts, and meaning that depends on position rather than text (which boxes an arrow connects, what changed between two diagram versions, exactly when a meeting starts and ends in a calendar screenshot). Also more reliable at multistep computer use from screenshots: at its default effort it matched the success rate Claude Opus 5 reached only at a much higher effort setting, in fewer steps and with roughly 40% fewer tokens.
+
+### Behavioral shifts (prompt-tunable)
+
+- **Re-evaluate Claude Opus 5-specific instructions.** Instructions tuned for Claude Opus 5's behavior (the verbosity, over-verification, and scope prompts under § Behavioral shifts of § Migrating to Claude Opus 5) may no longer be needed - keep them as the starting point, then re-test each on your own evals rather than carrying them over untouched.
+- **Progress updates:** covered above - render the `thinking` blocks, and ask in the system prompt for the cadence you want.
+- **Frontend design defaults:** asked for frontend work without design direction, it falls back on a few default styles, and a general instruction such as "avoid a generic AI look" mostly swaps one default for another. **It responds well to instructions that name specific patterns to avoid.** Work iteratively - look at which styles the first result used instead, and extend the list:
+
+ > *"Output a vanilla HTML/CSS personal website with placeholder data. Do not use a cream or off-white background, italic accent words in headlines, numbered "01/02/03" section labels, monospace labels, or pill-shaped buttons."*
+
+- **Ingesting complex visual inputs:** it reads charts, diagrams, and screenshots considerably more precisely out of the box, so **harness scaffolding built for visual inputs on earlier models may no longer be needed - re-test it.** For the densest inputs, two things still add accuracy: higher-resolution images (most of all for technical drawings), and image-processing tools - run it as an agent with a container holding the raw images and libraries such as PIL and OpenCV so it can crop, zoom, measure, and verify; if a container is too much overhead, a cropping tool alone still helps. It uses these tools more effectively at higher effort levels; without tools, raising effort improves its reading of technical drawings but does little for charts.
+- **Long turns:** at `xhigh` / `max`, turns run longer than on Claude Opus 5 - plan timeouts, streaming, and progress UX accordingly, and lower effort before prompting for brevity.
+
+### Claude Opus 5.5 Migration Checklist
+
+- [ ] **[BLOCKS]** Model ID -> `claude-opus-5-5` (Bedrock: `anthropic.claude-opus-5-5`). Coming from Opus 4.8 or older, the Claude Opus 5 checklist first - except that disabling thinking is not an option.
+- [ ] **[BLOCKS]** Remove `thinking: {type: "disabled"}` and `{type: "enabled", budget_tokens}` on every route - both 400 at every effort level. Choose an effort level instead; size `max_tokens` for thinking plus the reply; read content blocks by `type`; pass `thinking` blocks back unmodified.
+- [ ] **[BLOCKS]** Replace `tool_choice` `any` / `tool` with `auto` plus `strict: true` (steering in the prompt, and a check that the call happened) or structured outputs - on `count_tokens` and Batches too.
+- [ ] **[BLOCKS]** Computer use: declare `{"type": "computer_toolset_20260801"}` (no beta header, no `name` / display size) instead of `computer_20251124`, and update the agent loop for member `tool_use` blocks (action = block `name`), batch actions, and `toolset_name` on every result; confirm the toolset is offered on your platform. Test on Claude Opus 5 first.
+- [ ] **[BLOCKS]** If the harness builds `messages` itself: run the preserved-thinking three-step check (§ Migrating to Claude Fable 5.1 from Claude Fable 5) - accounts created on or after 2026-08-31 are enforced by default on every platform; set `prefix_mismatch_behavior` explicitly under `thinking-binding-controls-2026-08-01` and replace every history edit with its append-only form. Declare from the first request any tool the session may need later.
+- [ ] **[BLOCKS]** Handle `stop_reason: "refusal"` before reading `content` (new `bio` and `reasoning_extraction` categories) and ship a fallback opt-in; `reasoning_extraction` is not retried on a fallback.
+- [ ] **[TUNE]** Set `effort` explicitly - the default is `medium`, one level below Claude Opus 5's `high` - and re-run the sweep including `low` / `medium`; lower effort before adding "think less" prompts; reserve `xhigh` / `max` for measured gains; use per-message effort (beta) to vary it without a cache reset.
+- [ ] **[TUNE]** If the UI showed text between tool calls: `display: "updates"` (beta) or `"summarized"`, render non-empty `thinking` blocks; give the model a send-message tool (declared at session start) for verbatim mid-turn content; ask in the system prompt for the update cadence you want.
+- [ ] **[TUNE]** If a router or fallback can move the conversation to another model, expect it to run without Claude Opus 5.5's thinking blocks (only Claude Fable 5.1 / Claude Mythos 5.1 on the Claude API keep them).
+- [ ] **[TUNE]** If thinking was disabled on Claude Opus 5: start at `low`, remove reasoning-in-the-response instructions, re-test the thinking-disabled mitigation instruction, delete any "don't think" rule.
+- [ ] **[TUNE]** Re-test visual-input scaffolding (may be unnecessary now); for frontend work, name the specific default patterns to avoid rather than asking for "no generic look"; re-evaluate Claude Opus 5-specific verbosity / verification / scope instructions.
+- [ ] **[TUNE]** Re-baseline cost and latency at the chosen effort level: $4 / $20, cache reads $0.20 per MTok; separate rate-limit pool; no Priority Tier; fast mode is Claude API only at $8 / $40.
+
+---
+
+
## Verify the Migration
-After updating, spot-check that the new model is actually being used. Replace `YOUR_TARGET_MODEL` with the model string you migrated to (e.g. `claude-fable-5-1`, `claude-opus-5`, `claude-opus-4-8`, `claude-opus-4-7`, `claude-sonnet-5`, `claude-sonnet-4-6`, `claude-haiku-4-5`) and keep the assertion prefix in sync:
+After updating, spot-check that the new model is actually being used. Replace `YOUR_TARGET_MODEL` with the model string you migrated to (e.g. `claude-fable-5-1`, `claude-opus-5-5`, `claude-opus-5`, `claude-opus-4-8`, `claude-opus-4-7`, `claude-sonnet-5`, `claude-sonnet-4-6`, `claude-haiku-4-5`) and keep the assertion prefix in sync:
```python
YOUR_TARGET_MODEL = "claude-opus-5" # or "claude-opus-4-7", "claude-sonnet-5", "claude-sonnet-4-6", "claude-haiku-4-5"
diff --git a/skills/claude-api/shared/models.md b/skills/claude-api/shared/models.md
index cc12416..0228cd7 100644
--- a/skills/claude-api/shared/models.md
+++ b/skills/claude-api/shared/models.md
@@ -61,6 +61,7 @@ curl https://api.anthropic.com/v1/models/claude-opus-4-8 \
| Claude Mythos 5.1 | `claude-mythos-5-1` | - | 1M | 128K | Active (Project Glasswing only) |
| Claude Fable 5 | `claude-fable-5` | - | 1M | 128K | Active |
| Claude Mythos 5 | `claude-mythos-5` | - | 1M | 128K | Active (Project Glasswing only) |
+| Claude Opus 5.5 | `claude-opus-5-5` | - | 1M | 128K | Active (launching; use only when the user names it) |
| Claude Opus 5 | `claude-opus-5` | - | 1M | 128K | Active |
| Claude Opus 4.8 | `claude-opus-4-8` | - | 1M | 128K | Active |
| Claude Opus 4.7 | `claude-opus-4-7` | - | 1M | 128K | Active |
@@ -70,9 +71,10 @@ curl https://api.anthropic.com/v1/models/claude-opus-4-8 \
| Claude Haiku 4.5 | `claude-haiku-4-5` | `claude-haiku-4-5-20251001` | 200K | 64K | Active |
### Model Descriptions
-- **Claude Fable 5.1** - Anthropic's most capable widely released model, for the most demanding reasoning and long-horizon agentic work. Successor to Claude Fable 5 in the same tier at the same per-token price ($10/$50 per MTok; cache reads $0.25/MTok - 0.025x, a quarter of Claude Fable 5's; batch $5/$25); stronger long-running agentic coding, knowledge work with documents/spreadsheets/slides, multistep research, vision, long-context retrieval, and computer use. Same API surface as Claude Fable 5 (thinking always on, no prefill, no sampling params, `refusal` stop reason, 512-token cache minimum) with three breaking changes: forced tool use (`tool_choice` `any` / `tool`) returns a 400; thinking blocks are bound to the producing model (only Claude Mythos 5.1 can read them - other models drop them); and editing earlier turns invalidates thinking blocks ("preserved thinking"; new accounts created on/after 2026-08-31 get a 400 on edited history; later models enforce it for everyone; the opt-in controls are per-platform - `shared/platform-availability.md`). Adds per-message `effort`, turn-scoped `clear_at` system messages, `thinking.display: "updates"` progress updates, and content provenance. Same tokenizer as Claude Fable 5; 1M context (default), 128K max output. Covered Model: 30-day retention required (ZDR only if expressly authorized by Anthropic) - ZDR orgs get `400 invalid_request_error`, as on Claude Fable 5. No Priority Tier; shares the Fable 5.x rate-limit pool. See `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5.
+- **Claude Fable 5.1** - Anthropic's most capable widely released model, for the most demanding reasoning and long-horizon agentic work. Successor to Claude Fable 5 in the same tier at the same per-token price ($10/$50 per MTok; cache reads $0.25/MTok - 0.025x, a quarter of Claude Fable 5's; batch $5/$25); stronger long-running agentic coding, knowledge work with documents/spreadsheets/slides, multistep research, vision, long-context retrieval, and computer use. Same API surface as Claude Fable 5 (thinking always on, no prefill, no sampling params, `refusal` stop reason, 512-token cache minimum) with three breaking changes: forced tool use (`tool_choice` `any` / `tool`) returns a 400; thinking blocks are bound to the producing model (only Claude Mythos 5.1 can read them - other models drop them); and editing earlier turns invalidates thinking blocks ("preserved thinking"; new accounts created on/after 2026-08-31 get a 400 on edited history on every platform, and enforcement scope is decided per model; the opt-in controls are per-platform - `shared/platform-availability.md`). Adds per-message `effort`, turn-scoped `clear_at` system messages, `thinking.display: "updates"` progress updates, and content provenance. Same tokenizer as Claude Fable 5; 1M context (default), 128K max output. Covered Model: 30-day retention required (ZDR only if expressly authorized by Anthropic) - ZDR orgs get `400 invalid_request_error`, as on Claude Fable 5. No Priority Tier; shares the Fable 5.x rate-limit pool. See `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5.
- **Claude Fable 5** / **Claude Mythos 5** (`claude-fable-5` / `claude-mythos-5`) - the previous Fable / Mythos release: same tier, limits and per-token pricing as Claude Fable 5.1, which adds three breaking API changes over them (see above; cache reads here are $1/MTok rather than Claude Fable 5.1's $0.25); still served and selectable by id. Claude Mythos 5 ran no safety classifiers, so `stop_reason: "refusal"` does not occur on it. Prefer claude-fable-5-1 for new work.
- **Claude Mythos 5.1** - The same model as Claude Fable 5.1 (same capabilities, limits, per-token pricing, API behavior), offered only to approved Project Glasswing customers; successor to Claude Mythos 5 (which itself succeeded the invitation-only `claude-mythos-preview`). Unlike Claude Mythos 5 it runs safeguards that depend on the access program, so handle `stop_reason: "refusal"`. Not offered on Claude Platform on AWS. Use it only when the org participates in Project Glasswing; otherwise use `claude-fable-5-1`.
+- **Claude Opus 5.5** - Successor to Claude Opus 5 in the Opus line for long-running agentic coding and knowledge work, at a lower price ($4 / $20 per MTok; cache reads $0.20). Same 1M context, 128K output, tokenizer, and feature set as Claude Opus 5, with four breaking changes: thinking can't be disabled (effort is the only control, default `medium`), forced `tool_choice` 400s, thinking blocks are tied to the model and the conversation, and computer use needs the `computer_toolset_20260801` toolset. Broader safety classifiers (`bio` and `reasoning_extraction` join `cyber`). Use it only when the user names it until launch; see `shared/model-migration.md` -> Migrating to Claude Opus 5.5.
- **Claude Opus 5** - For complex agentic coding and enterprise work; a step-change over Claude Opus 4.8, strongest on deep reasoning, agentic and long-horizon work, and test-time compute scaling, at half the cost of Claude Fable 5.1 (Claude Fable 5.1 remains the highest-capability tier). Safety classifiers can return `stop_reason: "refusal"` - handle it before reading `content`. A drop-in upgrade at Opus 4.8's pricing ($5/$25 per MTok) with the same feature set. Thinking is on by default (omitting `thinking` runs adaptive; `{type: "adaptive"}` is equivalent), and `thinking: {type: "disabled"}` is available only at effort `high` or lower - pairing it with `xhigh`/`max` returns a 400. Raw thinking tokens are never returned. Full effort ladder through `max`; 512-token prompt-cache minimum (down from 1024 on Opus 4.8); fast mode on the Claude API only. Elevated cybersecurity safeguards. Separate rate-limit bucket from the combined Opus 4.x pool. 1M context window (default and maximum), 128K max output. See `shared/model-migration.md` -> Migrating to Claude Opus 5.
- **Claude Opus 4.8** - The most capable model in the Opus 4 series - highly autonomous, state-of-the-art on long-horizon agentic work, knowledge work, and memory; clearer, warmer writing. Same API surface as Opus 4.7 (adaptive thinking only; sampling parameters and `budget_tokens` removed). 1M context window at standard API pricing (no long-context premium). See `shared/model-migration.md` -> Migrating to Opus 4.8 - a 4.7 -> 4.8 move is a model-ID swap plus prompt re-tuning, no new breaking changes.
- **Claude Opus 4.7** - Previous-generation Opus. Highly autonomous; strong on long-horizon agentic work, knowledge work, vision, and memory. Adaptive thinking only; sampling parameters and `budget_tokens` removed. 1M context window. See `shared/model-migration.md` -> Migrating to Opus 4.7.
@@ -123,6 +125,7 @@ When a user asks for a model by name, use this table to find the correct model I
| "mythos preview" | `claude-mythos-5-1` (successor to `claude-mythos-preview` - see migration guide) |
| "opus" | `claude-opus-5` |
| "opus 5" | `claude-opus-5` |
+| "opus 5.5" | `claude-opus-5-5` (launching - see migration guide) |
| "opus 4.8" | `claude-opus-4-8` |
| "opus 4.7" | `claude-opus-4-7` |
| "opus 4.6" | `claude-opus-4-6` |
diff --git a/skills/claude-api/shared/platform-availability.md b/skills/claude-api/shared/platform-availability.md
index 57ec0cc..0a5c3d2 100644
--- a/skills/claude-api/shared/platform-availability.md
+++ b/skills/claude-api/shared/platform-availability.md
@@ -29,7 +29,7 @@ Columns: **1P** = first-party Claude API, **P-AWS** = Claude Platform on AWS (An
| &nbsp;&nbsp;Advisor tool | beta | beta | No | No | No | |
| **Client-implemented tools** | | | | | | |
| &nbsp;&nbsp;Bash, text editor, memory | Yes | Yes | Yes | Yes | Yes | |
-| &nbsp;&nbsp;Computer use | beta | beta | beta | beta | beta | |
+| &nbsp;&nbsp;Computer use | beta | beta | beta | beta | beta | `computer_20251124` and older versions: beta on all five platforms. Claude Opus 5.5 accepts only `computer_toolset_20260801`: GA on the Claude API and Google Cloud with no beta header, not offered on the other platforms - toolset availability is not final, confirm at launch (`shared/model-migration.md` -> Migrating to Claude Opus 5.5, breaking change 4) |
| **Agentic / orchestration** | | | | | | |
| &nbsp;&nbsp;Agent Skills (Messages API) | Yes | Yes | No | No | beta | Foundry: Hosted on Anthropic deployments only - Hosted on Azure returns a 400 |
| &nbsp;&nbsp;Programmatic tool calling | Yes | Yes | No | No | Yes | Foundry: Hosted on Anthropic deployments only - Hosted on Azure returns a 400 |
@@ -41,14 +41,14 @@ Columns: **1P** = first-party Claude API, **P-AWS** = Claude Platform on AWS (An
| &nbsp;&nbsp;Files API | Yes | Yes | No | No | beta | Foundry: Hosted on Anthropic deployments only - Hosted on Azure returns a 400 |
| &nbsp;&nbsp;Models API | Yes | Yes | No | No | No | |
| **Other** | | | | | | |
-| &nbsp;&nbsp;Mid-conversation system messages | Yes | Yes | Yes | Yes | No | Claude Opus 5, Claude Opus 4.8, Claude Fable 5, Claude Fable 5.1, Claude Mythos 5, Claude Mythos 5.1; not Claude Sonnet 5. Bedrock: InvokeModel passthrough, not ARN-versioned models |
+| &nbsp;&nbsp;Mid-conversation system messages | Yes | Yes | Yes | Yes | No | Claude Opus 5, Claude Opus 5.5, Claude Opus 4.8, Claude Fable 5, Claude Fable 5.1, Claude Mythos 5, Claude Mythos 5.1; not Claude Sonnet 5. Bedrock: InvokeModel passthrough, not ARN-versioned models |
| &nbsp;&nbsp;Mid-conversation tool changes | beta | beta | beta | beta | No | Same models as mid-conversation system messages; beta `mid-conversation-tool-changes-2026-07-01` |
| &nbsp;&nbsp;Turn-scoped (`clear_at`) system messages | beta | beta | beta | beta | No | Same models as mid-conversation system messages; beta `mid-conversation-system-clear-at-2026-08-21` (on Bedrock/Vertex pass the value as a beta) |
-| &nbsp;&nbsp;Per-message `effort` (system message `output_config`) | beta | No | No | No | No | Claude Fable 5.1, Claude Mythos 5.1, Claude Opus 5; beta `mid-conversation-output-config-2026-07-01`; Claude API at launch (Bedrock/Vertex/Foundry unconfirmed; Claude Opus 5 excluded on Bedrock) |
-| &nbsp;&nbsp;`thinking.display: "updates"` | beta | beta | beta | beta | beta | Claude Fable 5.1, Claude Mythos 5.1, Claude Fable 5; beta `thinking-display-updates-2026-08-18` (pass the beta value per platform); without it `"updates"` is rejected as an unknown `display` value |
+| &nbsp;&nbsp;Per-message `effort` (system message `output_config`) | beta | No | No | No | No | Claude Fable 5.1, Claude Mythos 5.1, Claude Opus 5, Claude Opus 5.5; beta `mid-conversation-output-config-2026-07-01`; Claude API at launch (Bedrock/Vertex/Foundry unconfirmed; Claude Opus 5 excluded on Bedrock) |
+| &nbsp;&nbsp;`thinking.display: "updates"` | beta | beta | beta | beta | beta | Claude Fable 5.1, Claude Mythos 5.1, Claude Fable 5, Claude Opus 5.5; beta `thinking-display-updates-2026-08-18` (pass the beta value per platform); without it `"updates"` is rejected as an unknown `display` value |
| &nbsp;&nbsp;Thinking block-binding controls | beta | beta | per model | per model | No | `thinking.block_binding` + `input_transformations`; beta `thinking-binding-controls-2026-08-01` (on Bedrock via the `anthropic_beta` body field); the controls beta arrives per model on Bedrock/Vertex - until then the header is rejected; the history-editing enforcement itself follows the account-age rule in `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5 |
| &nbsp;&nbsp;Server-side `fallbacks` | beta | beta | No | No | No | `"default"` -> beta `server-side-fallback-2026-07-01`; array form -> beta `server-side-fallback-2026-06-01` |
-| &nbsp;&nbsp;Fast mode | beta | No | No | No | No | Research preview, beta `fast-mode-2026-02-01`, first-party API only |
+| &nbsp;&nbsp;Fast mode | beta | No | No | No | No | Research preview, beta `fast-mode-2026-02-01`, first-party API only (Claude Opus 5 / Opus 4.8 at $10 / $50; Claude Opus 5.5 at $8 / $40 - its fast-mode docs flip after the model launch, confirm before quoting) |
| &nbsp;&nbsp;Cache diagnostics | beta | No | No | No | No | First-party API only |
| &nbsp;&nbsp;Task budgets | beta | beta | No | No | No | Beta header `task-budgets-2026-03-13`; 3P availability not documented - assume unsupported |
diff --git a/skills/claude-api/shared/prompt-audit.md b/skills/claude-api/shared/prompt-audit.md
index 7f8a35b..ff3835e 100644
--- a/skills/claude-api/shared/prompt-audit.md
+++ b/skills/claude-api/shared/prompt-audit.md
@@ -78,14 +78,15 @@ These aren't tuned down; they're swapped for the feature that replaced them. For
|---|---|
| "Think step by step", `<scratchpad>`/`<thinking>` tag instructions | Adaptive thinking (`thinking: {type: "adaptive"}`) + `effort`. On thinking models the incantation is redundant at best; control depth via configuration, not prose. |
| "Use the think tool to plan" / "plan before acting" | Delete - current models plan without being told, and these cause over-planning. If behavior is still too aggressive after cleanup, lower `effort` rather than adding prose. |
-| "Show your thinking" / required reasoning sections in the output | Read thinking blocks via the API. On Claude Fable 5.1, instructing reasoning reproduction can trigger a `refusal` (reasoning extraction) - this is an explicit audit item when migrating. |
+| Prose that steers thinking depth ("think harder", "think less", "don't overthink", "answer without deliberating"), and any rule telling the model not to think | `effort`. Where thinking is always on (Claude Fable 5.1, Claude Opus 5.5) effort is the only thinking control, and lowering it cuts thinking, cost, and latency more reliably than prose; a "don't think" rule can't be followed there. Keep a "reply directly" line only where a measurement on a latency-sensitive route shows it helps. |
+| "Show your thinking" / required reasoning sections in the output | Read thinking blocks via the API. On Claude Fable 5.1 and Claude Opus 5.5, instructing reasoning reproduction can trigger a `refusal` (reasoning extraction, not retried on a fallback) - this is an explicit audit item when migrating. |
| Assistant-turn prefill (`{"role": "assistant", "content": "{"`) and the JSON-forcing stack around it: stop-sequences, regex extraction, retry-on-parse loops, "output ONLY valid JSON" | Structured outputs (`output_config.format`). Prefill 400s on 4.6-and-later Opus- and Sonnet-tier models and Claude Fable 5.1 - confirm in the per-target section of `shared/model-migration.md` before claiming the error. Where it applies, the *surrounding code* is cruft too - audit the request builder, not just the prompt string. Only a **trailing** assistant turn is a prefill - partial or complete-looking (a few-shot block ending on the assistant side still counts): assistant turns mid-array are ordinary conversation history and must stay. |
| "Summarize progress every N tool calls" choreography; hard word caps (`at most N words`) | Delete and re-baseline: current models narrate appropriately, and output caps starve reasoning on hard problems. Prefer qualitative length guidance ("be concise") over numeric caps tuned against an older model's verbosity. |
| Inline lookup tables, point systems, arithmetic rubrics the model must compute | Data in files or tool results; arithmetic in code. Leave the model the judgment layer. |
| `budget_tokens`, non-default `temperature`/`top_p`/`top_k`, stale beta headers, dead 400-retry paths | See `shared/model-migration.md` - whether each one hard-errors or is merely deprecated depends on the target model, so take the error claim from the per-target section there, not from memory. Where it does error, the retry/workaround code around it is removable too. |
-| Forced tool use - `tool_choice: {type: "any"}` / `{type: "tool", name: ...}` - and the JSON-via-forced-tool pattern | Prompt instruction naming the tool under `tool_choice: auto` (steering), or structured outputs (extraction). Returns a 400 on Claude Fable 5.1 / Claude Mythos 5.1 (and Mythos Preview); elsewhere it works but is usually a prompt-instruction in disguise - `strict: true` keeps the schema guarantee under `auto`. Audit the retry-on-missing-tool loop around it as well. |
+| Forced tool use - `tool_choice: {type: "any"}` / `{type: "tool", name: ...}` - and the JSON-via-forced-tool pattern | Prompt instruction naming the tool under `tool_choice: auto` (steering), or structured outputs (extraction). Returns a 400 on Claude Fable 5.1 / Claude Mythos 5.1 / Claude Opus 5.5 (and Mythos Preview); elsewhere it works but is usually a prompt-instruction in disguise - `strict: true` keeps the schema guarantee under `auto`. Audit the retry-on-missing-tool loop around it as well. |
-**Signals:** `think step by step|take a deep breath`; `<scratchpad>|<thinking>` in instructions; `stop_sequences` guarding JSON; `json.loads` inside retry loops; `budget_tokens|temperature|top_p` in request code; `every \d+ (tool calls|messages)`; `at most \d+ (words|sentences)`.
+**Signals:** `think step by step|take a deep breath`; `think (harder|less)|don'?t overthink|do not think`; `<scratchpad>|<thinking>` in instructions; `stop_sequences` guarding JSON; `json.loads` inside retry loops; `budget_tokens|temperature|top_p` in request code; `every \d+ (tool calls|messages)`; `at most \d+ (words|sentences)`.
#### 1c. Over-specification - describe the goal, not the method
@@ -105,21 +106,25 @@ These aren't tuned down; they're swapped for the feature that replaced them. For
| Pattern | Why it's cruft now | Fix |
|---|---|---|
-| Model-version workarounds: formatting fixes, over-refusal softeners, retry hints, "known issue with [model]" comments, date-conditional guidance | Nobody owns the removal, so prompts accumulate the union of every generation's mitigations | Each mitigation names (or gets traced to) the model it patched; if that model is retired, remove and re-test |
+| Model-version workarounds: formatting fixes, over-refusal softeners, retry hints, "known issue with [model]" comments, date-conditional guidance | Nobody owns the removal, so prompts accumulate the union of every generation's mitigations | Each mitigation names (or gets traced to) the model it patched; if that model is retired, remove and re-test. Targeting Claude Opus 5.5, the verbosity, over-verification, and scope instructions written for Claude Opus 5 (`shared/model-migration.md` -> Migrating to Claude Opus 5 -> Behavioral shifts) are the named re-test candidates: keep them as the starting point and test each removal on your own evals |
+| Thinking-disabled mitigations: the combined "say a sentence before a tool call / say so if no tool fits / no internal XML tags" instruction, and reasoning-in-the-response substitutes for thinking | Written for Claude Opus 5 running with thinking off, where those artifacts appear; on Claude Opus 5.5 thinking is always on and the instruction may be dead weight (the reasoning-substitute half can also be declined as `reasoning_extraction`) | Re-test and remove what no longer reproduces; read reasoning from `display: "summarized"` blocks (see `shared/model-migration.md` -> Migrating to Claude Opus 5.5 -> Prompts written for thinking disabled) |
+| Visual-input scaffolding: step-by-step chart-reading instructions, OCR or table-extraction pre-passes, mandatory crop or zoom steps | Built for weaker vision; Claude Opus 5.5 reads charts, diagrams, and screenshots considerably more precisely without tools | Remove one piece at a time and re-test on your own images. Keep image-processing tools (crop, zoom, measure in a container) and higher-resolution inputs for the densest material - they still add accuracy, most of all on technical drawings - so this is a finding about prompt text and pre-passes, not about tools |
| Migration-relative phrasing: "X now works differently", "also counts", "no longer" | The text is a diff against a previous prompt version the model never saw; relative phrasing implies phantom alternatives | Write as if current rules are the only rules that ever existed |
| Patch accretion: many narrow conditionals, each traceable to one incident | The model navigates a maze of special cases instead of a coherent principle, and fails unpredictably between them; an eval win for adding a line on top of the stack is not evidence the stack should exist | Generalize the principle or fix the underlying context; test removals, not just additions |
| Unenforced instructions: rules no code path, eval, or reviewer checks - visibly violated in the app's own transcripts | If nothing checks it and nobody noticed, it carries no signal - and behavioral rules that could be hooks, allowlists, or schema validators are less reliable as prose | Enforce in code what can be enforced in code; delete what nothing enforces and nobody misses |
| Identity stubs standing in for context ("You are a helpful assistant") | A role line is fine as a one-sentence focus-setter; the defect is an identity statement *substituting* for audience, product, and quality bar | Don't flag a short role line; flag when it's the only context the prompt gives |
-| Update suppressors written for chatty models: "hold all findings for the final response", "don't narrate", "no interim updates" | Tuned against models that over-narrated; current models (Claude Fable 5.1 especially) under-narrate with these present, and the harness may not be requesting the model's between-tool progress notes at all (`thinking.display: "updates"`) | Remove first and re-test; if more narration is still wanted, replace with a specific line saying *when* user-facing text is wanted (see `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5 -> User-facing progress updates) |
+| Update suppressors written for chatty models: "hold all findings for the final response", "don't narrate", "no interim updates" | Tuned against models that over-narrated; current models (Claude Fable 5.1 and Claude Opus 5.5 especially) under-narrate with these present, and the harness may not be requesting the model's between-tool progress notes at all - on both they come back as `thinking` blocks (`thinking.display: "updates"`), so a client that renders only `text` blocks looks silent | Remove first and re-test; if more narration is still wanted, replace with a specific line saying *when* user-facing text is wanted (see `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5 -> User-facing progress updates, and Migrating to Claude Opus 5.5 -> Text between tool calls comes back in thinking blocks) |
| Anti-formatting rules: "never use bullets", "no headers", "no bold" | Written against models that over-formatted; Claude Fable 5.1 already under-formats, so the rule now strips formatting the reader wanted | Remove, or replace with a rule that says when formatting is appropriate (the conditional-formatting snippet in the Claude Fable 5.1 migration section) |
| Instruction re-insertion every few turns ("reminder: ..." repeated on a cadence in the harness) | A retention crutch for models that lost instructions over long sessions; current models retain a once-stated instruction, and each repeat costs tokens and, under preserved thinking's history-editing check, is a history edit if it is later removed | Remove the repetition and re-test; where a genuinely per-turn reminder remains, send it as a turn-scoped (`clear_at`) system message - or a text block after the tool results - and never delete earlier copies |
-**Signals:** retired model names in prompts or comments (`claude-2|claude-3|claude-instant|3\.5|3\.7`); `hold (all )?(findings|results)|don't narrate|no interim`; `never use (bullets|headers|bold)|no (bullet|header)`; `reminder:` on a turn cadence; `before|after [date]` conditionals; `now|no longer|instead of` attached to behavioral rules; rules whose reason nobody remembers; `^You are (a|an) (helpful|expert)` with nothing task-specific following.
+**Signals:** retired model names in prompts or comments (`claude-2|claude-3|claude-instant|3\.5|3\.7`); `hold (all )?(findings|results)|don't narrate|no interim`; `internal (XML )?tags|before (each|a) tool call` mitigations; `OCR|crop|zoom|axis labels` in image-reading prompts; `never use (bullets|headers|bold)|no (bullet|header)`; `reminder:` on a turn cadence; `before|after [date]` conditionals; `now|no longer|instead of` attached to behavioral rules; rules whose reason nobody remembers; `^You are (a|an) (helpful|expert)` with nothing task-specific following.
#### 1e. Prohibition clusters - judge by provenance, not by whether the model "needs it"
A run of unconditional "never / don't / must not" lines is audited by asking, for each, **does it carry a stated reason or encode a real business/policy constraint?** - not "does the target model still need this guardrail?" (the latter question keeps everything, because nothing is *harmful* to say). Prohibitions that encode observable constraints (refund caps, data rules, compliance language, promises the business must not make) stay, ideally with their reason beside them. Prohibitions that merely describe an undesirable *output style* with no provenance - banned phrases, tic lists, "don't start with 'Certainly'" written against an older model's habits - are cruft: restate the desired style positively in one line, or attach the real reason if there is one. A surrounding cluster of legitimate reasoned prohibitions does not launder the no-provenance ones mixed into it; classify each line separately.
+**One exception to "restate positively": frontend design direction on Claude Opus 5.5.** Asked for frontend work without direction, it falls back on a few default styles, and a general "avoid a generic AI look" mostly swaps one default for another - that vague line is the finding. A list that names the specific defaults to avoid (a cream background, italic accent words in headlines, numbered "01/02/03" section labels, monospace labels, pill-shaped buttons) is the form that works: keep it, and extend it from the styles the first result used instead. Propose rewriting the vague line into named patterns, never deleting the list.
+
#### 1f. Output-shaping choreography - one pattern, remove every limb
Fixed interim-update cadences ("after every third tool call, post a progress note"), numeric output ceilings ("under 120 words", "at most five bullets"), and cut-the-detail instructions are manifestations of the **same** over-constraint pattern, written for models that padded or rambled. They are removed *together*: a stated operational reason ("queue throughput", "supervisors skim") does not convert a numeric clamp into a keeper - re-express the goal as audience/outcome framing without the number ("replies are scan-able and answer only what was asked"), and keep any genuinely format-sensitive requirement as a format instruction, not a word count. Removing the cadence while keeping the ceilings leaves the pattern in place.
@@ -162,6 +167,8 @@ Skill files (`SKILL.md`, `CLAUDE.md`, rule files) inherit everything in Group 1,
The same audit keeps surfacing these next to prompt cruft; report them even though they're not prompt text.
- **API fossils**: parameters and headers that error or are deprecated on the target model - the per-model lists live in `shared/model-migration.md`; treat each migration checklist as a removal checklist.
+- **Thinking config and `max_tokens` sized for the wrong model**: `thinking: {type: "disabled"}` and `budget_tokens` 400 on Claude Opus 5.5 (thinking is always on - remove the field and set `effort`, default `medium`), and a `max_tokens` sized for a thinking-off route cuts replies off, because thinking counts toward it even when its text isn't returned (64K is a reasonable start for long agentic coding turns).
+- **History-editing harness**: request-building code that rewrites the `system` prompt, the `tools` array, or earlier messages mid-session - including deleting a per-turn reminder, or adding a tool late (declare any tool the session may need, such as a send-the-user-a-message tool, from the first request) - invalidates preserved-thinking blocks on Claude Fable 5.1 and Claude Opus 5.5. Report each edit site with its append-only replacement from `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5.
- **Cache-hostile ordering**: timestamps, UUIDs, per-user content interpolated above stable content. Read `shared/prompt-caching.md` -> Silent invalidators, and run its greps during this audit.
- **Budget countdowns rendered into context**: surfacing remaining-token counts to the model can cause premature wrap-up behavior; avoid showing them where possible.
- **An LLM executor for a deterministic plan**: agent sessions whose transcript is the same loop body N times; calls whose inputs fully determine outputs. **Run this check, don't wait to notice it**: in every pipeline, batch job, or agent loop, *count the model-call sites* and ask of each whether its inputs fully determine its output. Routing, tallying, normalizing, filtering, and formatting steps go back into plain code; keep exactly one model call where the work is genuinely adaptive (classifying the ambiguous remainder, writing the judgment summary). Zero model calls is an over-fix when a judgment step exists - name the one call that stays.
diff --git a/skills/claude-api/shared/prompt-caching.md b/skills/claude-api/shared/prompt-caching.md
index f5b3304..9027059 100644
--- a/skills/claude-api/shared/prompt-caching.md
+++ b/skills/claude-api/shared/prompt-caching.md
@@ -233,11 +233,11 @@ Implication: you can change `tool_choice` per-request without losing the tools+s
| System prompt content | A `{"role": "system", "content": "..."}` message - see § Mid-conversation system messages above | Claude Opus 5, Claude Opus 4.8, Claude Fable 5, Claude Fable 5.1, Claude Mythos 5, Claude Mythos 5.1 - **already available today**, no beta header |
| Per-turn reminder (inject, then delete next request) | A turn-scoped `clear_at: "next_user_message"` system message, left in the transcript - see § Mid-conversation system messages above (without the beta: a text block after the `tool_result` blocks, earlier copies kept) | Same models as mid-conversation system messages, behind `mid-conversation-system-clear-at-2026-08-21` |
| `effort` change | A `{"role": "system", "content": [], "output_config": {"effort": ...}}` message - see `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5 | Claude Fable 5.1, Claude Mythos 5.1, Claude Opus 5, behind `mid-conversation-output-config-2026-07-01` |
-| Dropped thinking blocks (a Claude Fable 5.1 / Claude Mythos 5.1 block replayed to a model that can't read it, or a history-editing-check `drop_block`) | None - the API drops the block on that request and the messages cache changes from its position onward; tools and system caches are intact. Blocks the receiving model can read, passed back unchanged, keep the cache intact | - |
+| Dropped thinking blocks (a Claude Fable 5.1 / Claude Mythos 5.1 / Claude Opus 5.5 block replayed to a model that can't read it - only Claude Fable 5.1 / Claude Mythos 5.1 on the Claude API read Claude Opus 5.5's - or a history-editing-check `drop_block`) | None - the API drops the block on that request and the messages cache changes from its position onward; tools and system caches are intact. Blocks the receiving model can read, passed back unchanged, keep the cache intact | - |
Model switch has no escape hatch: caches are model-scoped. Keep the main loop on one model and spawn a subagent for cheaper sub-tasks (see `agent-design.md` § Caching for Agents).
-**Thinking blocks and the messages cache (model-specific).** On Claude Fable 5, Claude Fable 5.1, Claude Mythos 5, Claude Mythos 5.1, Mythos Preview, Opus 4.5 and later, and Sonnet 4.6 and later, previous-turn thinking blocks are preserved by default, so passing a regular (non-tool-result) user message with thinking enabled leaves the messages cache valid. On earlier Opus and Sonnet models and all Haiku models through Haiku 4.5, that same request strips previously-cached thinking blocks from context, and every message after the first stripped block falls out of cache - in an agent loop this shows up as a `cache_creation_input_tokens` spike on turns where a plain user message follows tool use. (Toggling thinking on/off between requests is a separate, all-models invalidator of the messages cache - see the hierarchy table above. Changing `output_config.effort` behaves the same as changing thinking parameters; setting the model's default effort explicitly is equivalent to omitting it, so pinning the default costs nothing.)
+**Thinking blocks and the messages cache (model-specific).** On Claude Fable 5, Claude Fable 5.1, Claude Mythos 5, Claude Mythos 5.1, Mythos Preview, Opus 4.5 and later (Claude Opus 5.5 included), and Sonnet 4.6 and later, previous-turn thinking blocks are preserved by default, so passing a regular (non-tool-result) user message with thinking enabled leaves the messages cache valid. On earlier Opus and Sonnet models and all Haiku models through Haiku 4.5, that same request strips previously-cached thinking blocks from context, and every message after the first stripped block falls out of cache - in an agent loop this shows up as a `cache_creation_input_tokens` spike on turns where a plain user message follows tool use. (Toggling thinking on/off between requests is a separate, all-models invalidator of the messages cache - see the hierarchy table above. Changing `output_config.effort` behaves the same as changing thinking parameters; setting the model's default effort explicitly is equivalent to omitting it, so pinning the default costs nothing.)
---
diff --git a/skills/claude-api/shared/tool-use-concepts.md b/skills/claude-api/shared/tool-use-concepts.md
index 04b1bd0..5d67654 100644
--- a/skills/claude-api/shared/tool-use-concepts.md
+++ b/skills/claude-api/shared/tool-use-concepts.md
@@ -103,7 +103,7 @@ Control when Claude uses tools:
Any `tool_choice` value can also include `"disable_parallel_tool_use": true` to force Claude to use at most one tool per response. By default, Claude may request multiple tool calls in a single response.
-**Claude Fable 5.1, Claude Mythos 5.1, and Mythos Preview reject forced tool use:** `{"type": "any"}` and `{"type": "tool", "name": ...}` return a 400 there (`tool_choice: type "tool" and "any" are not supported for this model.` - on `count_tokens` and Batches too). It is a model-specific restriction (Claude Fable 5 and Claude Opus 5 accept them). Use `{"type": "auto"}` and state the expectation in the prompt ("Use the get_weather tool to answer") - `strict: true` on the tool keeps the schema-valid-arguments guarantee `any` gave you - or structured outputs (`output_config.format`) when the forced call only existed to extract JSON. `auto` and `none` are unaffected; `disable_parallel_tool_use` with `auto` still means at most one call (the "exactly one" combination with `any`/`tool` is gone). Combining `tool_choice` `any` with `strict: true` applies only on models that support forced tool use. See `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5.
+**Claude Fable 5.1, Claude Mythos 5.1, Claude Opus 5.5, and Mythos Preview reject forced tool use:** `{"type": "any"}` and `{"type": "tool", "name": ...}` return a 400 there (`tool_choice: type "tool" and "any" are not supported for this model.` - on `count_tokens` and Batches too). It is a model-specific restriction (Claude Fable 5 and Claude Opus 5 accept them). Because `auto` does not guarantee a call, check that one was made and retry if it wasn't. Use `{"type": "auto"}` and state the expectation in the prompt ("Use the get_weather tool to answer") - `strict: true` on the tool keeps the schema-valid-arguments guarantee `any` gave you - or structured outputs (`output_config.format`) when the forced call only existed to extract JSON. `auto` and `none` are unaffected; `disable_parallel_tool_use` with `auto` still means at most one call (the "exactly one" combination with `any`/`tool` is gone). Combining `tool_choice` `any` with `strict: true` applies only on models that support forced tool use. See `shared/model-migration.md` -> Migrating to Claude Fable 5.1 from Claude Fable 5.
---
@@ -387,7 +387,9 @@ For full documentation, use WebFetch:
Computer use lets Claude interact with a desktop environment (screenshots, mouse, keyboard). It is a client-side tool - your application provides the environment and executes the actions Claude requests; Anthropic processes the screenshots and action requests in real time but does not host the environment or retain the data.
-For full documentation, use WebFetch:
+**Two request shapes.** The current one is the **computer toolset** - GA on the Claude API and Google Cloud, no beta header: one `tools` entry `{"type": "computer_toolset_20260801"}` with **no `name`** and no display dimensions, plus an optional `configs` map to turn member tools off (`{"zoom": {"enabled": false}}`; all 17 members, `zoom` included, are on by default). Claude's calls are `tool_use` blocks whose `name` is the member (`screenshot`, `left_click`, `type`, `zoom`, ...) carrying `"toolset_name": "computer"`, often several per turn; return one `tool_result` per call in the next `user` message, **each echoing `"toolset_name": "computer"`** (only `screenshot` / `zoom` need an image; `OK` suffices for the rest). Coordinates are in the pixel space of the full screenshots you return, also after a `zoom`, and screenshots must already fit the model's image limits. The earlier `computer_20251124` tool (beta `computer-use-2025-11-24`, a `name: "computer"` entry with `display_width_px` / `display_height_px`, actions in `input.action`) keeps working on the models and platforms that offer it - Bedrock, Claude Platform on AWS, and Foundry offer only the earlier beta versions today - and the two forms can't share a request. **Claude Opus 5.5 accepts only the toolset**: `computer_20251124` returns a 400 there (`shared/model-migration.md` -> Migrating to Claude Opus 5.5 -> Breaking change 4 has the request and agent-loop changes; test them on Claude Opus 5, which accepts both).
+
+For full documentation (member reference, batch actions, scaling, the `computer_20251124` migration steps), use WebFetch:
- URL: `https://platform.claude.com/docs/en/agents-and-tools/computer-use/overview`