skill-evolutionlisted

# skill-evolution skill-evolution reads io-trace JSONL and produces evolution candidate reports for repeated tool, workflow, and sequence patterns. If a session-intent ledger is provided, it uses the intent summary to explain why the tool sequence occurred. This skill is report-only. It does not edit skills, install hooks, promote memory, or create new automation. ## Contents - [When To Use](#when-to-use) - [Conduct Feedback](#conduct-feedback) - [Procedure](#procedure) - [Output Format](#output-format) - [Quality Decisions](#quality-decisions) - [Warnings](#warnings) ## When To Use - Analyze `~/.ghost-alice/io-trace.jsonl` or isolated fixtures. - Find repeated tool sequences in a project. - Interpret repeated behavior with `.tmp/session-intent/<platform>/<session-id>/intent-state.json`. - Answer when the user asks whether there are skill upgrades, update recommendations, or evolution backlog items to review. - Gather evidence before proposing a new skill, hook, or doctor check. ## Conduct Feedback Tool-frequency, workflow, and sequence mining finds automation candidates, but it does not show whether the agent's behavior matched what the user asked. `conduct_feedback` fills that gap. Read it from the session-intent ledger as a first-class evolution signal. Open `conduct_feedback` entries form the update-recommendation backlog. `occurrence_count` preserves repeated same-id corrections inside one session, so chronic completion or workflow failures do not collapse i