Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the user asks about Codex Rules behavior, injected project rules, supported rule file locations, matching, or environment configuration.
.claude/skills/code-yeongyu-rules/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -27% | 0% |
Codex Rules is automatic once the plugin is enabled. It injects:
SessionStart and UserPromptSubmitapply_patch by defaultDynamic PostToolUse output is injected as additional context and is deduplicated per plugin data session. Codex Rules does not rewrite tool output.
Supported project sources:
CONTEXT.md.omo/rules/**/*.md.claude/rules/**/*.md.cursor/rules/**/*.md.github/instructions/**/*.md.github/copilot-instructions.mdSupported environment knobs:
CODEX_RULES_DISABLED=1CODEX_RULES_MODE=both|static|dynamic|offCODEX_RULES_MAX_RULE_CHARS=<number>CODEX_RULES_MAX_RESULT_CHARS=<number>CODEX_RULES_ENABLED_SOURCES=CONTEXT.md,.omo/rulesThe legacy PI_RULES_* variables are accepted as fallbacks for users migrating from pi-rules.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 2,431 | 2,372 | -2% | 1 | 1 | 0% | 427 | 679 | +59% | 0 | 0 | — |
case-01 | fail→pass | 10,782 | 10,175 | -6% | 1 | 1 | 0% | 2,448 | 2,658 | +9% | 0 | 0 | — |
case-02 | fail→pass | 14,217 | 10,800 | -24% | 1 | 1 | 0% | 2,703 | 2,547 | -6% | 0 | 0 | — |
case-03 | fail→pass | 6,978 | 5,077 | -27% | 1 | 1 | 0% | 1,496 | 1,419 | -5% | 0 | 0 | — |
case-04 | pass→pass | 9,970 | 5,620 | -44% | 1 | 1 | 0% | 1,693 | 1,459 | -14% | 0 | 0 | — |
case-05 | pass→pass | 2,069 | 1,558 | -25% | 1 | 1 | 0% | 394 | 524 | +33% | 0 | 0 | — |
case-07 | fail→pass | 27,949 | 9,182 | -67% | 1 | 1 | 0% | 846 | 714 | -16% | 0 | 0 | — |
case-08 | pass→pass | 7,561 | 2,336 | -69% | 1 | 1 | 0% | 1,543 | 749 | -51% | 0 | 0 | — |
case-09 | fail→pass | 10,372 | 11,156 | +8% | 1 | 1 | 0% | 782 | 570 | -27% | 0 | 0 | — |
case-10 | pass→pass | 8,102 | 2,216 | -73% | 1 | 1 | 0% | 1,529 | 701 | -54% | 0 | 0 | — |
case-15 | pass→pass | 5,458 | 1,291 | -76% | 1 | 1 | 0% | 1,069 | 481 | -55% | 0 | 0 | — |
case-11 | pass→pass | 11,028 | 2,611 | -76% | 1 | 1 | 0% | 2,037 | 771 | -62% | 0 | 0 | — |
case-12 | fail→pass | 12,119 | 3,578 | -70% | 1 | 1 | 0% | 2,111 | 848 | -60% | 0 | 0 | — |
case-13 | pass→pass | 6,598 | 1,298 | -80% | 1 | 1 | 0% | 1,236 | 496 | -60% | 0 | 0 | — |
case-14 | fail→pass | 5,355 | 1,529 | -71% | 1 | 1 | 0% | 1,096 | 545 | -50% | 0 | 0 | — |
case-16 | pass→pass | 10,615 | 1,645 | -85% | 1 | 1 | 0% | 2,209 | 576 | -74% | 0 | 0 | — |
case-17 | fail→pass | 7,130 | 1,869 | -74% | 1 | 1 | 0% | 1,442 | 622 | -57% | 0 | 0 | — |
case-18 | fail→pass | 8,269 | 2,461 | -70% | 1 | 1 | 0% | 1,681 | 748 | -56% | 0 | 0 | — |
case-19 | pass→pass | 6,875 | 1,441 | -79% | 1 | 1 | 0% | 1,296 | 542 | -58% | 0 | 0 | — |
case-20 | fail→pass | 9,292 | 1,370 | -85% | 1 | 1 | 0% | 1,467 | 499 | -66% | 0 | 0 | — |
case-21 | fail→pass | 7,787 | 1,625 | -79% | 1 | 1 | 0% | 1,229 | 599 | -51% | 0 | 0 | — |
case-22 | fail→pass | 7,739 | 1,707 | -78% | 1 | 1 | 0% | 1,576 | 567 | -64% | 0 | 0 | — |
case-23 | fail→pass | 11,770 | 2,795 | -76% | 1 | 1 | 0% | 2,166 | 871 | -60% | 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. 23 cases were attempted, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +57 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.