Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Evaluate Claude Code rules in .claude/rules/. Use for frontmatter, globs, and quality audits.
.claude/skills/athola-rules-eval/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 131% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 9% | 0% |
abstract:skills-eval)abstract:hooks-eval)hookify:writing-rules)This skill evaluates Claude Code rules in .claude/rules/ directories against quality standards. It validates YAML frontmatter, glob pattern syntax, content quality, and directory organization. Rules files support path-scoped conditional loading via paths frontmatter and unconditional rules (no paths field).
Key validations: YAML syntax errors, unquoted glob patterns, Cursor-specific fields (alwaysApply, globs), overly broad patterns, content verbosity, and naming conventions.
bash# Evaluate rules in current project /rules-eval # Evaluate specific directory /rules-eval .claude/rules/ # Detailed analysis with recommendations /rules-eval --detailed
.claude/rules/ for all .md files (including subdirectories)| Category | Points | Focus | |----------|--------|-------| | Frontmatter Validity | 25 | YAML syntax, required fields, correct field names | | Glob Pattern Quality | 20 | Syntax, specificity, quoting | | Content Quality | 25 | Actionable, concise, non-conflicting | | Organization | 15 | Naming, structure, symlink usage | | Token Efficiency | 15 | Rule size, redundancy detection |
| Score | Level | |-------|-------| | 91-100 | Excellent - Production-ready | | 76-90 | Good - Minor improvements possible | | 51-75 | Basic - Needs optimization | | 26-50 | Below Standards - Significant issues | | 0-25 | Critical - Invalid or broken rules |
modules/frontmatter-validation.mdmodules/glob-pattern-analysis.mdmodules/content-quality-metrics.mdmodules/organization-patterns.mdscripts/rules_validator.pyabstract:skills-eval - Skill evaluation frameworkabstract:hooks-eval - Hook evaluation framework.md file under .claude/rules/ (including subdirectories) receives a quality score(0-100) with per-category breakdown across the five dimensions.
before any score is reported.
alwaysApply, globs) are flagged asnon-compliant with the expected Claude Code schema.
corrective action, not just a score.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 10,053 | 5,365 | -47% | 1 | 1 | 0% | 1,612 | 1,001 | -38% | 0 | 0 | — |
case-02 | fail→fail | 2,711 | 5,209 | +92% | 1 | 1 | 0% | 301 | 1,143 | +280% | 0 | 0 | — |
case-03 | pass→fail | 15,029 | 3,093 | -79% | 1 | 1 | 0% | 2,281 | 1,120 | -51% | 0 | 0 | — |
case-04 | pass→pass | 5,762 | 11,899 | +107% | 1 | 1 | 0% | 902 | 2,747 | +205% | 0 | 0 | — |
case-05 | fail→pass | 7,102 | 10,372 | +46% | 1 | 1 | 0% | 1,065 | 2,464 | +131% | 0 | 0 | — |
case-06 | pass→pass | 7,538 | 6,819 | -10% | 1 | 1 | 0% | 1,190 | 1,892 | +59% | 0 | 0 | — |
case-07 | fail→pass | 8,081 | 6,574 | -19% | 1 | 1 | 0% | 1,306 | 1,812 | +39% | 0 | 0 | — |
case-08 | pass→pass | 9,457 | 8,905 | -6% | 1 | 1 | 0% | 1,466 | 2,220 | +51% | 0 | 0 | — |
case-09 | fail→pass | 9,701 | 4,288 | -56% | 1 | 1 | 0% | 1,757 | 1,499 | -15% | 0 | 0 | — |
case-10 | pass→pass | 4,562 | 3,355 | -26% | 1 | 1 | 0% | 758 | 1,339 | +77% | 0 | 0 | — |
case-11 | pass→pass | 8,680 | 3,322 | -62% | 1 | 1 | 0% | 1,227 | 1,287 | +5% | 0 | 0 | — |
case-12 | pass→pass | 9,643 | 3,394 | -65% | 1 | 1 | 0% | 1,617 | 1,361 | -16% | 0 | 0 | — |
case-13 | fail→pass | 14,998 | 7,290 | -51% | 1 | 1 | 0% | 2,230 | 2,075 | -7% | 0 | 0 | — |
case-14 | fail→pass | 7,098 | 2,550 | -64% | 1 | 1 | 0% | 1,101 | 1,202 | +9% | 0 | 0 | — |
case-15 | pass→pass | 10,328 | 4,758 | -54% | 1 | 1 | 0% | 1,638 | 1,516 | -7% | 0 | 0 | — |
case-16 | pass→pass | 6,503 | 6,031 | -7% | 1 | 1 | 0% | 1,073 | 1,786 | +66% | 0 | 0 | — |
case-17 | pass→pass | 9,691 | 4,799 | -50% | 1 | 1 | 0% | 1,366 | 1,717 | +26% | 0 | 0 | — |
case-18 | pass→pass | 11,832 | 1,903 | -84% | 1 | 1 | 0% | 1,710 | 1,091 | -36% | 0 | 0 | — |
case-19 | pass→pass | 4,712 | 2,911 | -38% | 1 | 1 | 0% | 697 | 1,243 | +78% | 0 | 0 | — |
case-20 | pass→pass | 5,589 | 4,429 | -21% | 1 | 1 | 0% | 944 | 1,490 | +58% | 0 | 0 | — |
case-21 | pass→pass | 8,169 | 7,097 | -13% | 1 | 1 | 0% | 1,234 | 1,940 | +57% | 0 | 0 | — |
case-22 | pass→pass | 8,366 | 5,620 | -33% | 1 | 1 | 0% | 1,269 | 1,612 | +27% | 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, and 20 counted toward the lift figure. The other 2 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 +18 percentage points is the difference between those two pass rates over the 20 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.