| name | autoresearch |
| description | Durable research-mission mode for long-running discovery, optimization, and evidence gathering under .omx/. |
Autoresearch
Purpose
Use $autoresearch when the work is not ready for direct implementation and needs a mission loop with durable artifacts, logs, and a launch boundary.
Use When
- you need to investigate patterns before committing to an implementation
- you want a durable optimization or discovery pass
- there is a clear mission, evaluator, or keep policy to refine
Execution Policy
- Define the mission in plain language before launching anything.
- Treat launch readiness as a gate, not a vibe.
- Keep the log useful: phase changes, findings, blockers, next decision.
- Write artifacts the next pass can consume without rereading the whole chat.
Workflow
- Clarify the mission:
- topic
- why it matters
- what counts as a good result
- what should be kept or discarded
- If intent is still fuzzy, route through
$deep-interview first.
- Initialize the mission:
omx autoresearch init <mission>
- Record the canonical artifacts:
.omx/research/summary.md
.omx/plans/<mission>-requirements.md
.omx/logs/autoresearch.log
- Log every real state transition with
omx autoresearch log.
- Hand off to
$plan, $team, or direct execution only when the mission is decision-ready.
Deliverables
- a clear mission statement
- research summary with findings and open questions
- explicit launch / refine-further decision
Final Check
- the mission is concrete
- the logs are current
- the next decision is obvious