Review my recent work from the past 30 days, or the full available record if there is less than 30 days of history, and find recurring manual workflows that are good candidates to package.
Prioritize evidence in this order:
- Recent Codex sessions and task summaries.
- Codex Memories and rollout summaries, especially where the same pattern shows up across more than one session.
- Chronicle, if it is available, to discover repeated work outside Codex. Use Chronicle only as a discovery layer; verify anything important in the appropriate source system when possible.
- Existing skills, custom agents, and automations, so you improve or reuse what already exists instead of creating duplicates.
Search broadly for work that is repeated, slow, error-prone, context-heavy, or better handled through a consistent process. Include workflows from coding, research, writing, planning, communication, operations, analysis, and personal administration.
Only move a candidate forward when it meets these conditions:
- it has happened at least twice, or it is clearly likely to happen again and would be costly to repeat manually;
- it has stable inputs, a repeatable process, and a clear output or stopping point;
- packaging it would meaningfully improve speed, quality, consistency, or reliability;
- it is not already covered well enough by an existing asset.
Choose the smallest useful package:
- Skill: a reusable workflow, checklist, or playbook.
- Custom subagent: a focused specialist role or bounded investigation task that is useful to delegate.
- Automation: a scheduled or recurring check, report, reminder, or monitor.
- Extend existing: improve something that already covers most of the need.
- Skip: leave it alone if the work is too one-off, vague, sensitive, poorly evidenced, or already handled.
Start by producing a compact shortlist. For each item, include:
- the recurring workflow;
- supporting evidence and dates;
- frequency and confidence;
- recommended form: skill, subagent, automation, extend existing, or skip;
- why it is or is not worth packaging.
After the shortlist, create only the high-confidence missing items. Keep each one narrow, practical, source-aware, and easy to validate. Do not create speculative, overlapping, or overly broad assets.
Finish with:
- what you created or extended;
- what you intentionally skipped;
- what needs more evidence before it should be packaged.