recall

name	recall
description	Recall prior decisions and evidence with state-first search and bounded history fallback.

Run an OmA recall search to recover prior decisions without replaying full chat history.

Input: $ARGUMENTS

Intent:

• Find relevant prior context quickly with minimal token/runtime cost.
• Prefer compact OmA state artifacts before scanning broader history.

Protocol:

1. Parse arguments into:
   • query (required free text; if missing, return usage + examples)
   • optional since (examples: 48h, 7d, 2026-03-10)
   • optional limit (default 6, hard cap 12)
   • optional scope (state, recent, all; default state)
2. Search in this order and stop early when limit is reached:
   • .omg/state/taskboard.md
   • .omg/state/checkpoint.md
   • .omg/state/workflow.md
   • .omg/state/workspace.json
   • .omg/state/doctor.md
   • .omg/state/notify.json
3. If matches are insufficient and scope is recent or all, run bounded fallback:
   • inspect only recent transcript/log files (newest first)
   • max 20 files, max 200 matched lines total
   • skip binary files, vendor directories, and .git
4. Deduplicate near-identical snippets and keep each snippet under about 180 characters.
5. For each match, capture source path and a lightweight locator (line number or key).
6. If filesystem tools are available, write .omg/state/recall.md with query, filters, and top matches.
7. Suggest the next best resume command (/oma:taskboard, /oma:checkpoint, /oma:team-verify, etc.) based on findings.

Output format:

Recall Summary

• query:
• scope:
• since:
• searched:
• matched:

Top Matches

1. source: ... evidence: ...
2. ...

Next Command

• ...