workflow-auditlisted

# workflow-audit — Mine the agent-chat corpus for optimization targets Find 10–15 high-leverage interventions in HOW the user works with agents, ranked by `(bucket_coverage × cost)`, output as a single PII-scrubbed markdown report. This skill is intended to be **dispatched as a sub-agent** so the caller's context stays clean and the analysis is reproducible across machines. ## When to use - User asks to "audit my chat history", "find optimization targets", "what am I correcting agents on repeatedly", or similar. - User explicitly invokes `/workflow-audit`. - Periodically (e.g. monthly) as a maintenance ritual on the user's CC setup. ## Arguments `$ARGUMENTS` — optional, positional: 1. Output path. Default: `~/workflow-audit-<YYYYMMDD>.md`. 2. Weeks of corpus to scan. Default: `12`. ## Prerequisites The agent must have: 1. **`mache` binary** for the chat-corpus ingest. If missing: ```bash git clone https://github.com/agentic-research/mache ~/.cache/workflow-audit/mache cd ~/.cache/workflow-audit/mache && task install ``` 2. **`chat-embed` binary** from ley-line-open. If missing: ```bash git clone https://github.com/agentic-research/ley-line-open ~/.cache/workflow-audit/llo cd ~/.cache/workflow-audit/llo && cargo install --path rs/ll-open/chat-embed ``` 3. **Python 3** with stdlib only (used for JSONL parsing + scoring). No third-party deps. If any prereq install fails, **bail with a clear error**, do not fabricate findings. ## Load-bea