Skill Quality Auditor

Skill Quality Auditor

Evaluate, maintain, and improve skill quality with 9-dimension framework scoring.

Quick Start

# Single skill
skill-auditor evaluate <domain>/<skill-name> --json --store

# Batch with grade gate
skill-auditor batch <skill1> <skill2> --fail-below B --store

When to Use

• Evaluate skills before merge or publication using 9-dimension scoring
• Generate remediation plans, detect duplication (High at 20%, Critical at 35%), or enforce CI quality gates
• Validate eval scenario coverage and artifact conventions

When Not to Use

• Write the skill first — do not audit an unfinished draft
• Avoid using this as a substitute for peer review of logic or domain accuracy

Prerequisites

• MUST have the skill directory with SKILL.md and at least one eval scenario.
• MUST run --store at least once before remediate or trend can produce output.

Workflow

1. Run skill-auditor evaluate <skill> --json --store
2. Check artifacts and eval coverage using deterministic criteria
3. Generate a remediation plan with T-shirt sizing and score delta estimates
4. Re-run the auditor to verify improvement; if below target, focus on the lowest-scoring dimension

Anti-Patterns

NEVER skip baseline comparison in recurring audits. WHY: Score regressions go undetected without a prior stored audit. BAD: skill-auditor evaluate my-skill with no prior --store run. GOOD: skill-auditor evaluate my-skill --store after a prior stored audit exists.

NEVER ignore Knowledge Delta below 15/20. WHY: Low D1 means the skill adds no value over LLM baseline knowledge. BAD: Shipping a skill that restates generic framework documentation the LLM already knows. GOOD: Ensuring every skill section contains constraints or thresholds absent from public docs.

NEVER apply subjective scoring. WHY: Scores drift between evaluators and cannot be automated in CI pipelines. BAD: Assigning D6 a score without checking hard:soft marker ratios. GOOD: Running skill-auditor evaluate and using the numeric output as the canonical score.

NEVER create kitchen-sink skills covering unrelated tasks. WHY: Broad scope kills D7 pattern recognition and prevents correct agent triggering.

NEVER use harness-specific paths in skill content. WHY: Absolute paths break portability; use agent-neutral relative paths instead.

NEVER list references without "When to Use" conditions. WHY: Unconditional loading bloats context and penalises D5 progressive disclosure.

See Detailed Anti-Patterns for full failure modes, agent name references, and D4 heading rules.

Examples

Remediation workflow:

./dist/skill-auditor evaluate cmd/assets --json --store
# Score increases after each remediation cycle

PR-scoped triage:

# Extract changed skills from the PR diff and batch-audit them
./dist/skill-auditor batch cmd/assets testdata/fixtures/skill-full --fail-below B --store

Audit all skills:

./dist/skill-auditor batch $(find skills -name "SKILL.md" | sed 's|skills/||;s|/SKILL.md||' | tr '\n' ' ')

Troubleshooting

• A non-zero exit from batch means a skill scored below --fail-below; duplication exits 2 when a Critical (>=35%) pair is found.
• If a command exits below threshold, consider running evaluate --store to capture diagnostics; see Scripts Workflow for per-command failure modes.

Self-Audit

./dist/skill-auditor evaluate cmd/assets --json --store
# Expected: A grade (>= 126/140)

./dist/skill-auditor evaluate cmd/assets --json | jq '.grade'
# Confirms the grade: "A"

## References

### Framework

| Topic | Reference | When to Use |
| --- | --- | --- |
| Per-dimension criteria and bonus rules | [Dimensions](references/framework-dimensions.md) | Evaluating any dimension or understanding the rubric; skip if you only need the final grade |
| Score thresholds and grade bands | [Scoring Rubric](references/framework-scoring-rubric.md) | Calculating a total score or assigning a grade |
| A-grade checklist and red flags | [Quality Standards](references/framework-quality-standards.md) | Targeting A-grade or reviewing blockers |
| Trigger pattern density and keyword analysis | [Pattern Recognition](references/advanced-pattern-recognition.md) | Scoring D7 or improving description keywords |
| Canonical SKILL.md structure and References table standard | [SKILL Template](references/skill-template.md) | Authoring or refactoring a skill |

### Operations

| Topic | Reference | When to Use |
| --- | --- | --- |
| CI gate configuration and batch pass/fail logic | [Quality Thresholds](references/quality-thresholds-scoring.md) | Setting up CI quality gates |
| NEVER/WHY/BAD/GOOD failure modes per dimension | [Anti-Patterns](references/detailed-anti-patterns.md) | Explaining low scores or writing remediation guidance |
| T-shirt sizing and remediation roadmaps | [Remediation Planning](references/remediation-planning.md) | Writing a remediation plan for a C/D-grade skill |
| Deduplication workflow and aggregation guidance | [Duplication Detection](references/duplication-detection-algorithm.md) | Detecting skill overlap or planning aggregations |
| `skill-auditor evaluate/batch` usage and output formats | [Scripts Workflow](references/scripts-audit-workflow.md) | Running audits from the command line |
| Registry publication gates and tessl compliance checks | [Tessl Compliance](references/tessl-compliance-framework.md) | Preparing a skill for public registry submission |