What problem does it solve? Turning observed workflow patterns into actual Claude Code configuration is manual and error-prone. This Skill takes the candidates proposed by skill-miner and converts them into concrete artifacts—CLAUDE.md rule additions, skill scaffold contexts, hook configurations, and agent definitions—with approval flows and decision logging. ## Core Features & Use Cases - Kind-based dispatch: Routes each proposal candidate by suggested_kind to the right action path—immediate apply for CLAUDE.md, scaffold handoff for skills, guided creation for hooks and agents. - Safe CLAUDE.md apply: Runs security and conciseness pre-checks, shows a diff preview, and only then appends the rule to the appropriate section. - Guided hook/agent generation: Presents a design draft, waits for explicit user approval, then writes .claude/settings.json hook entries or .claude/agents/*.md files; cross-repo candidates produce handoff JSON instead. - Decision writeback: Records adopt/defer/reject outcomes with completion state to the decision log via skill_miner_decision.py and skill_miner_proposal.py. - Use Case: After a daytrace-session surfaces a proposal like "run lint before every git commit", ask to apply it and the Skill will draft the hook, get your approval, write the script and settings entry, and log the decision. ## Quick Start Ask the assistant to apply the selected proposal from skill-miner, for example by saying "apply proposal 1 as a hook".
