feat: skill matching protocol with intent classification#12
feat: skill matching protocol with intent classification#12
Conversation
…idence scoring Replace the upstream using-superpowers 1% rule with a local skill-matching skill that classifies user intent (QUESTION/CONTINUATION/ACTION), scores skill relevance with weighted signals, and responds with tiered transparency (auto-invoke >80, propose 30-80, skip <30). Restructure CLAUDE.md into System Behaviors and Personal Preferences sections, adding the skill matching override and skill transparency protocol. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5478e81149
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| ## Node.js / NVM | ||
| ### TDD Skill Override | ||
|
|
||
| For TDD guidance, use the local `skills/tdd/SKILL.md` skill. Ignore the superpowers test-driven-development skill — the local version is a comprehensive merge that supersedes it. |
There was a problem hiding this comment.
This override now instructs agents to use skills/tdd/SKILL.md, but that file is not present in the repo, so any workflow that tries to apply the TDD override cannot load the referenced guidance and will fall back unpredictably (or fail to apply TDD rules entirely). I checked the current skills/**/SKILL.md paths and there is no skills/tdd/SKILL.md, so this new reference is broken as committed.
Useful? React with 👍 / 👎.
1 participant
Summary
skills/skill-matching/SKILL.md— replaces the upstreamusing-superpowers1% rule with intent classification (QUESTION/CONTINUATION/ACTION), confidence scoring with weighted signals, and tiered response (auto-invoke >80, propose 30-80, skip <30)CLAUDE.mdinto System Behaviors and Personal Preferences sections, adding skill matching override and skill transparency protocolTest plan
🤖 Generated with Claude Code