Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →General discipline rules for coding agents. Use when any agent performs coding, debugging, refactoring, architecture changes, repository edits, test writing or running, tool-driven implementation, or code review. Enforces query-before-guessing, ask-before-ambiguous-execution, human confirmation for business assumptions, reuse before new interfaces, active validation, architecture compliance, honest uncertainty, and cautious refactoring. Chinese triggers: 认知纪律, 八荣八耻, 荣耻, 编码纪律, 主动测试, 遵循规范.
.claude/skills/cercatrovato-agent-rongchi/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 386% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -16% | 0% |
Apply this skill as a compact behavioral discipline for coding-agent work. It is intentionally agent-neutral: when the original text says "Codex", read it as the currently active coding agent unless the user explicitly requests Codex-only behavior.
Before changing code, invoking tools, or claiming completion:
Preserve this original Chinese block verbatim when adapting the skill to other agent protocols:
textCodex八荣八耻： 以瞎猜接口为耻，以认真查询为荣； 以模糊执行为耻，以寻求确认为荣； 以臆想业务为耻，以人类确认为荣； 以创造接口为耻，以复用现有为荣； 以跳过验证为耻，以主动测试为荣； 以破坏架构为耻，以遵循规范为荣； 以假装理解为耻，以诚实无知为荣； 以盲目修改为耻，以谨慎重构为荣。
Use careful lookup over interface guessing. Use confirmation over vague execution. Use human or repository evidence over imagined business logic. Use existing interfaces over invented ones. Use active testing over skipped validation. Use architecture compliance over structural damage. Use honest uncertainty over false understanding. Use cautious refactoring over blind modification.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 8,467 | 3,850 | -55% | 1 | 1 | 0% | 1,511 | 1,189 | -21% | 0 | 0 | — |
case-01 | fail→fail | 18,214 | 14,016 | -23% | 1 | 1 | 0% | 3,989 | 3,477 | -13% | 0 | 0 | — |
case-02 | fail→pass | 15,104 | 6,882 | -54% | 1 | 1 | 0% | 3,403 | 1,905 | -44% | 0 | 0 | — |
case-03 | fail→pass | 11,166 | 11,306 | +1% | 1 | 1 | 0% | 2,483 | 2,870 | +16% | 0 | 0 | — |
case-04 | fail→pass | 15,256 | 8,937 | -41% | 1 | 1 | 0% | 3,310 | 2,175 | -34% | 0 | 0 | — |
case-05 | pass→pass | 6,275 | 8,386 | +34% | 1 | 1 | 0% | 1,203 | 2,103 | +75% | 0 | 0 | — |
case-07 | fail→pass | 2,724 | 11,138 | +309% | 1 | 1 | 0% | 503 | 2,447 | +386% | 0 | 0 | — |
case-08 | fail→fail | 17,183 | 12,947 | -25% | 1 | 1 | 0% | 3,118 | 2,764 | -11% | 0 | 0 | — |
case-09 | pass→pass | 3,433 | 3,604 | +5% | 1 | 1 | 0% | 549 | 980 | +79% | 0 | 0 | — |
case-10 | fail→fail | 2,843 | 2,199 | -23% | 1 | 1 | 0% | 469 | 866 | +85% | 0 | 0 | — |
case-11 | pass→fail | 4,260 | 2,072 | -51% | 1 | 1 | 0% | 664 | 753 | +13% | 0 | 0 | — |
case-12 | fail→fail | 9,976 | 7,461 | -25% | 1 | 1 | 0% | 2,137 | 1,959 | -8% | 0 | 0 | — |
case-13 | pass→pass | 8,080 | 7,483 | -7% | 1 | 1 | 0% | 1,462 | 1,752 | +20% | 0 | 0 | — |
case-14 | pass→pass | 7,806 | 5,507 | -29% | 1 | 1 | 0% | 1,758 | 1,662 | -5% | 0 | 0 | — |
case-20 | pass→pass | 1,819 | 1,757 | -3% | 1 | 1 | 0% | 320 | 784 | +145% | 0 | 0 | — |
case-15 | fail→fail | 6,010 | 2,559 | -57% | 1 | 1 | 0% | 1,296 | 961 | -26% | 0 | 0 | — |
case-16 | pass→pass | 5,340 | 10,264 | +92% | 1 | 1 | 0% | 918 | 2,113 | +130% | 0 | 0 | — |
case-17 | pass→pass | 8,335 | 2,873 | -66% | 1 | 1 | 0% | 1,611 | 998 | -38% | 0 | 0 | — |
case-18 | pass→pass | 9,903 | 8,933 | -10% | 1 | 1 | 0% | 2,224 | 2,104 | -5% | 0 | 0 | — |
case-19 | fail→pass | 10,264 | 6,084 | -41% | 1 | 1 | 0% | 1,860 | 1,568 | -16% | 0 | 0 | — |
case-21 | pass→pass | 2,621 | 2,558 | -2% | 1 | 1 | 0% | 474 | 975 | +106% | 0 | 0 | — |
case-22 | pass→pass | 2,026 | 1,665 | -18% | 1 | 1 | 0% | 305 | 802 | +163% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +18 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.