EmmanuelRTM /investigate — Stage 1: Problem Investigator
Arguments: $ARGUMENTS
Current state: !python3 .claude/scripts/ledger.py status
- Resolve the run:
--run <id>given → useworkspace/runs/<id>.- A topic string given → scaffold a fresh run first:
python3 .claude/scripts/new_run.py $ARGUMENTS. - Neither → use the ACTIVE_RUN shown above; if none, ask for a topic.
- Launch the
problem-investigatorsubagent with the run directory, topic, and offline flag fromrun-config.json. It follows its own three-round protocol and the relevance gate. - On
ok: false(gate abort): relay the reason; suggest better seeds inworkspace/seeds/or a broader topic. - On success: run
python3 .claude/scripts/bib_validate.py <run>(must pass), append ledger event{"event":"stage_investigate","detail":"complete"}, and show the user the brief's section headers plus the source count.
Next stage: /discover.