Turn an opportunity into a plan that is "something you can start today + can validate at low cost that you can truly make money."
Cross-platform note (one-time, not repeated below)
Below, "web search/open webpage": Claude Code uses WebSearch / WebFetch, Codex uses its built-in equivalent capabilities.allowed-tools is a Claude Code permission declaration; Codex ignores it.
Prerequisite check
- Read
.money-state.jsonto find this opportunity. If it has not been verified or is judged high-risk ❌,first urge the user to go throughmoney-verify, and do not directly make a plan. - For uncertain ⚠️ ones, a plan can be made, but the plan must take "validate whether the first payment can be received" as its first goal.
The plan must include 5 parts
- Validation first step (most important): one minimal action, with the goal of using as little time/money as possible to confirm "someone will really pay for this". Example: first list a minimal service on Xianyu, first post 3 pieces of content to see the data, first take 1 smallest job.Do not invest a lot of time before validation.
- Action checklist for the first two weeks: specific to "what to do every day/every week", quantifiable (how many posts, how many applications, how many orders taken).
- Honest income expectations: roughly how long until the first payment in the early stage, and the amount range. State clearly whether this is "trading time for small money + building a portfolio" or quick money. Do not draw a pie in the sky.
- Stop-loss line: what conditions mean you should stop (example: after investing X hours there is zero positive feedback / any anti-fraud red line appears / payment is required).
- Reusable AI efficiency points: in this opportunity, exactly which step AI helps you improve efficiency at (do not let "using AI" stop at a slogan).
When needed, use web search to check the current specific operating path for this opportunity (how to activate the platform, how to take orders, how to withdraw).
Save to disk (including "prediction" — this is the starting point of the calibration loop)
Write the plan into .money-state.json 's active(chosen_id / started_at=today / plan), and change the opportunity status to chosen。
At the same time, write down active.prediction: estimated hours to invest, how long until the first payment, and the first income range (recorded_at=today).
Once this prediction is written, do not change it — later money-retro will reconcile "actual" against it, which is the only way to calibrate whether this method is accurate.
Wrap-up
Tip: after starting execution, use money-status to check progress; when the validation milestone or T+N days arrives, use money-retro to review actual investment and income.
Once again emphasize the stop-loss line and anti-fraud red lines — once the other party starts asking for money, stop immediately.