Read-only, changes nothing. Summarizes current progress and gives next-step suggestions.
Cross-platform note
allowed-tools is the Claude Code permission declaration; Codex can ignore it and use the local file tools available in the current session to accomplish the equivalent action.
Process
Step 1 — Read status
test -f .money-state.json && echo OK || echo MISSING
- MISSING → prompt to run
money-init。 - OK → read and display.
Step 2 — Display
- Profile + tier: current
tier(T0–T3) and rationale / skills / weekly available time / startup capital / region / face reveal. - Accumulated experience: if there is
lessons.md, summarize a few "works for me / pitfalls experienced firsthand" entries. - Candidate opportunities: list grouped by feasible ✅ / questionable ⚠️ / high-risk ❌, each marked with "where the money comes from".
- Currently executing: chosen opportunity + start date + whether the plan's first validation step has been completed.
- Review records: retro_log summary (investment/income/conclusion/lessons).
Step 3 — Suggest next steps
Intelligently route based on status:
- No candidates → "help me find opportunities" (money-find)
- Questionable opportunities not yet clarified → "verify XX" (money-verify)
- Feasible opportunities not yet chosen → "I want to do XX, help me make a plan" (money-plan)
- Opportunities being executed → reached validation checkpoint/T+N days → "review XX" (money-retro)
End with a sobering line from rubric E, reminding that the moat comes from skill accumulation, not the project itself.