feat(skill-review): strict-mode prompt + raise min_tool_calls 3 to 5

The review agent was creating too many low-value skills. Three changes
tighten the funnel:

- review_worker prompt: rewritten in "保守模式". Default decision is
  do-nothing. To create, all five AND conditions must hold (>=5 tool
  calls with step dependencies, an explicit fail->fix->success episode,
  cross-scenario reusability, no >=70% overlap with skill_list, can
  produce >=300 chars of substance). Adds 7 explicit NO categories
  (single-tool loops, one-off queries, LLM common knowledge, etc.)
  and a 4-question self-check before any create. Hard cap of 1 new
  skill per review, with patch preferred over create.
- min_tool_calls: 3 -> 5. Matches the prompt's ">=5 steps" hard
  condition so trivial turns never reach the LLM reviewer.
- Updated config.yaml, the default-value assertion in
  test_skill_review_worker, and the middleware docstring's "(default
  3)" annotation to keep all three references in sync.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
xuhdandClaude Opus 4.7 committed 2026-06-09 16:00:29 +08:00
1 parent 539a8bcec8
commit c493639119
5 files changed
+114 -77

No files matched your search

@@ -33,7 +33,7 @@ class SkillReviewMiddleware(AgentMiddleware[SkillReviewMiddlewareState]):
"""Dispatch a background skill review after every substantial agent turn.
'Substantial' is defined by ``skill_evolution.review_worker.min_tool_calls``
(default 3). Turns below the threshold are silently skipped to avoid
(default 5). Turns below the threshold are silently skipped to avoid
wasting LLM tokens on trivial exchanges.
"""