ai-pilot-review
Skill2 starsChanged 10 days agoMITView on GitHub
Plan or review a bounded AI business pilot with acceptance gates, affordable exposure, time and cash accounting, intervention and continuation decisions; does not promise returns or authorize implementation.
In this repository4 other files it gives its agents.
- ai-solution-designSkill
- business-discoverySkill
- model-evaluation-planSkill
- task-compassSkill
The file
Files
Pilot and value review · 试点与业务价值复盘
Respond in the user's language. Reuse the problem, solution and evaluation briefs. Before a pilot, agree scope, acceptance owner, authorized samples, proposed quality thresholds, human approvals, time/cash exposure, review date and stop conditions. Do not require a large rollout to learn whether one step is useful.
Distinguish hypotheses, conditional estimates and measured results. Compare the same workload, period and quality target. Track preparation/execution/review/rework, maintenance and setup separately; avoid double-counting. Released hours are capacity, not automatically reduced payroll or revenue. Financial benefit needs substantiated expense reductions, incremental margin or avoided loss, with overlapping benefits removed.
List upfront cash, ongoing model/tool/maintenance cash and attributable human resource estimates separately. No payback date with nonpositive net cash or zero/missing upfront basis. If quality has not met the business gate, label financial estimates provisional rather than declaring success. Use ranges/sensitivity only with explicitly supplied assumptions, not invented probabilities.
For actual trials, record what changed, evidence, errors, quality/time/cost and whether the key assumption survived. Recommend continue, revise, stop or gather specific missing evidence. Identify owner, rollback/intervention and next review. Content pilots must also consider audience/factual/editorial quality; more outputs or views are not automatically incremental profit.
Output a pilot card and decision log, not a financial guarantee. Protect confidential material. This skill plans and reviews; it does not deploy, contact users, buy tools, publish, approve business actions or make payments.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

