summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorAddy Osmani <addyosmani@gmail.com>2026-11-18 12:04:37 -0800
committerGitHub <noreply@github.com>2026-11-18 12:04:37 -0800
commit49d7b464f38c585449f12414008b40904986976f (patch)
tree24554777d6c06916a3c51884fad9dc0b901bbd76
parent9a69de588142a63c726b347d31a918d415a91fce (diff)
parentec0a2c83b72bfe5049e9dab1d7290e7f1b1fcfe1 (diff)
Merge #798: make the Gemini subagent example loadable
docs(per-agent-config): make the Gemini subagent example loadable
-rw-r--r--docs/advanced-per-agent-configuration.md11
1 files changed, 6 insertions, 5 deletions
diff --git a/docs/advanced-per-agent-configuration.md b/docs/advanced-per-agent-configuration.md
index d06a87b..c1a79b2 100644
--- a/docs/advanced-per-agent-configuration.md
+++ b/docs/advanced-per-agent-configuration.md
@@ -35,22 +35,23 @@ These are Claude Code conventions. Keep the shared, published `SKILL.md` limited
Gemini CLI and Antigravity discover skills the same way: they match `name` and `description`, then load the full skill on demand. Runtime orchestration lives in a separate resource, not in the skill frontmatter.
-- **Subagent definitions** live in `.gemini/agents/*.md`, which are distinct from skills. Their schema is `kind` (`local` or `remote`), `model`, `temperature`, `max_turns`, and `tools`. Note that this is a subagent schema, not a skill schema: the two should not be merged into one `SKILL.md`.
+- **Subagent definitions** live in `.gemini/agents/*.md`, which are distinct from skills. Their schema requires `name` and `description`, and accepts `kind` (`local` or `remote`), `model`, `temperature`, `max_turns`, and `tools`. Note that this is a subagent schema, not a skill schema: the two should not be merged into one `SKILL.md`.
- **Model routing** can send lighter tasks (formatting, documentation, git operations) to a faster model and reserve a stronger model for cognitively demanding work (debugging, security auditing, interface design).
- **Tool restrictions** enforce least-privilege per subagent, which also trims unused tool schemas from the prompt.
- **Turn limits** (`max_turns`) cap execution to avoid runaway correction loops.
### Model guidance
-Google recommends omitting `temperature` for Gemini 3.x models and using `thinking_level` instead. Do not hardcode `temperature` on skills or subagents routed to 3.x models; prefer `thinking_level` and revisit this guidance as the models evolve.
+Google recommends omitting `temperature` for Gemini 3.x models and using `thinking_level` instead. Do not hardcode `temperature` on skills or subagents routed to 3.x models, and revisit this guidance as the models evolve. `thinking_level` is a model setting, not a subagent field: set it in Gemini CLI's model configuration, not in the agent file.
```yaml
# .gemini/agents/security-auditor.md
---
+name: security-auditor
+description: Finds security vulnerabilities in code.
kind: local
-model: gemini-3-pro
-thinking_level: high # not: temperature: 0.1
-tools: [read_file, grep]
+model: gemini-3-pro # no temperature for 3.x models
+tools: [read_file, grep_search]
max_turns: 10
---
```