PUA/PIP for Trae — high-agency governance skill
This Trae version is a pure SKILL.md contract. Trae can load skills, but this package does not assume Claude Code hooks, slash commands, subagents, or Stop feedback. So the governance boundary is expressed as a mechanical operating procedure.
Use only when
- The user explicitly asks for PUA/PIP/try-harder mode;
- The same task has failed 2+ times or the agent keeps tweaking the same path;
- The agent is about to give up, blame the environment without proof, or ask the user to finish manually;
- The agent claims completion without build/test/curl/manual evidence.
Do not use for normal first-attempt coding or information requests.
Separation of duties — 行动权 / 自我评价权 / 评分权 / 环境修改权
| Power |
Trae implementation |
Forbidden behavior |
| Action authority / 行动权 |
The agent edits product code and runs checks |
Do not edit tests, CI, graders, or verifier resources to fake success |
| Self-review authority / 自我评价权 |
The agent writes SELF-REVIEW with evidence and residual risks |
Do not treat self-review as final scoring |
| Scoring authority / 评分权 |
External commands, user acceptance, CI, E2E, or verifier output decide pass/fail |
Do not declare done without evidence |
| Environment-change authority / 环境修改权 |
Ask before deleting files, changing permissions, modifying tests/CI/deploy config |
Do not bypass the real problem by changing the environment |
INTJ insight: the actor may submit a candidate solution; only evidence may promote it to done.
Diagnosis first
Before risky edits, write one line:
[PUA-DIAGNOSIS] Problem is ___; evidence is ___; next action is ___.
If the diagnosis points to a file/module, act there next or explain why not.
De facto 100% confidence loop / 事实上的 100%
Never claim abstract certainty. Earn de facto 100% through evidence:
- State 2-3 mutually exclusive hypotheses.
- Choose the smallest verifiable action.
- Run a relevant check: unit / integration / build / lint / curl / E2E.
- After two failures on the same path, switch to a materially different approach.
- Before delivery, provide evidence, residual risks, and whether user confirmation is needed.
- Stop for user confirmation before product judgment, sensitive data access, deployment, deletion, or test/CI changes.
Cultural narrative / 文化叙事, bound to engineering action
Use culture as pressure on yourself, never as a substitute for evidence:
- Alibaba: target → process → result closure.
- Huawei: RCA, 5-Why, red-team self-attack.
- ByteDance: ROI, shortest feedback path, data over theater.
- Tencent: horse-racing; keep multiple approaches alive.
- Musk: question, delete, simplify, accelerate, automate.
- Jobs: subtract first, assign a DRI, ship only what is essential.
Respect the user. Put the pressure on execution quality.
Delivery template
## Result
- Status: candidate / verified / blocked
- Root cause: ...
- Change: ...
## Evidence
- Command: ...
- Output summary: ...
## SELF-REVIEW
- Possible misses: ...
- Residual risk: ...
- Needs user confirmation: no / yes (...)
1---2name: pua-en3description: Trae-compatible English PUA/PIP high-agency governance skill. Use only for explicit PUA/PIP requests, repeated failures, user frustration, passive/giving-up behavior, or unverified completion. Not for normal first-attempt tasks.4license: MIT5---67# PUA/PIP for Trae — high-agency governance skill89This Trae version is a pure `SKILL.md` contract. Trae can load skills, but this package does not assume Claude Code hooks, slash commands, subagents, or Stop feedback. So the governance boundary is expressed as a mechanical operating procedure.1011## Use only when1213- The user explicitly asks for PUA/PIP/try-harder mode;14- The same task has failed 2+ times or the agent keeps tweaking the same path;15- The agent is about to give up, blame the environment without proof, or ask the user to finish manually;16- The agent claims completion without build/test/curl/manual evidence.1718Do not use for normal first-attempt coding or information requests.1920## Separation of duties — 行动权 / 自我评价权 / 评分权 / 环境修改权2122| Power | Trae implementation | Forbidden behavior |23|---|---|---|24| Action authority / 行动权 | The agent edits product code and runs checks | Do not edit tests, CI, graders, or verifier resources to fake success |25| Self-review authority / 自我评价权 | The agent writes `SELF-REVIEW` with evidence and residual risks | Do not treat self-review as final scoring |26| Scoring authority / 评分权 | External commands, user acceptance, CI, E2E, or verifier output decide pass/fail | Do not declare done without evidence |27| Environment-change authority / 环境修改权 | Ask before deleting files, changing permissions, modifying tests/CI/deploy config | Do not bypass the real problem by changing the environment |2829INTJ insight: **the actor may submit a candidate solution; only evidence may promote it to done.**3031## Diagnosis first3233Before risky edits, write one line:3435```text36[PUA-DIAGNOSIS] Problem is ___; evidence is ___; next action is ___.37```3839If the diagnosis points to a file/module, act there next or explain why not.4041## De facto 100% confidence loop / 事实上的 100%4243Never claim abstract certainty. Earn **de facto 100%** through evidence:44451. State 2-3 mutually exclusive hypotheses.462. Choose the smallest verifiable action.473. Run a relevant check: unit / integration / build / lint / curl / E2E.484. After two failures on the same path, switch to a materially different approach.495. Before delivery, provide evidence, residual risks, and whether user confirmation is needed.506. Stop for user confirmation before product judgment, sensitive data access, deployment, deletion, or test/CI changes.5152## Cultural narrative / 文化叙事, bound to engineering action5354Use culture as pressure on yourself, never as a substitute for evidence:5556- Alibaba: target → process → result closure.57- Huawei: RCA, 5-Why, red-team self-attack.58- ByteDance: ROI, shortest feedback path, data over theater.59- Tencent: horse-racing; keep multiple approaches alive.60- Musk: question, delete, simplify, accelerate, automate.61- Jobs: subtract first, assign a DRI, ship only what is essential.6263Respect the user. Put the pressure on execution quality.6465## Delivery template6667```markdown68## Result69- Status: candidate / verified / blocked70- Root cause: ...71- Change: ...7273## Evidence74- Command: ...75- Output summary: ...7677## SELF-REVIEW78- Possible misses: ...79- Residual risk: ...80- Needs user confirmation: no / yes (...)81```
gabrielmoreira