qms-audit-expert

I took a look at your QMS audit skill and noticed the spec compliance gaps are pulling the grade down—specifically around the progressive disclosure structure and how the reference materials are organized. The rubric seems solid for evaluation, but the skill's architecture feels like it's doing too much upfront before guiding users through the audit workflow step by step.

Links:

• GitHub
• Your agentic skill in the SkillzWave marketplace

The TL;DR

You're at 62/100, solidly in D territory. Based on Anthropic's 5-pillar rubric for agentic skills, you've got strong fundamentals (Spec Compliance at 12/15), but Utility and Progressive Disclosure Architecture are dragging things down. The biggest miss: all 9 referenced files are empty placeholders, which tanks your Utility score from what could be 18-20 down to 10.

What's Working Well

• Trigger phrases are solid — "ISO 13485", "internal audit", "CAPA verification" map directly to real use cases
• Frontmatter is clean — Valid YAML, correct naming conventions (hyphen-case)
• Good structural intent — The framework of "Audit Preparation → Execution → Nonconformity Management" makes sense as a workflow

The Big One: Empty Reference Files Kill Your Utility Score

Here's the thing: Your SKILL.md references 5 reference files and 4 scripts, but only 3 placeholder files exist with zero actual content. This is a massive missed opportunity.

Why it matters: Users invoke this skill expecting actionable guides (like iso13485-audit-guide.md with actual audit checklists), templates, and scripts. Instead they get pointers to empty files. That tanks Utility from potential 18/20 down to 10/20.

The fix:

• Create references/iso13485-audit-guide.md — Provide clause-by-clause audit methodology (ISO 13485:8.2), sample audit questions, and red flags
• Create references/nonconformity-severity-matrix.md — Actual risk/severity classification with examples
• Create scripts/audit-schedule-optimizer.py — Real Python code that takes process list + criticality scores and outputs an audit schedule
• Populate assets/ — Add real audit checklists, inspection templates, and CAPA tracking templates

Impact: +7 points minimum.

Other Things Worth Fixing

1. Add a Table of Contents — Your SKILL.md is 239 lines but lacks a TOC. Files >100 lines need navigation. Add it after the frontmatter.

2. Kill the marketing language — Phrases like "Expert-level", "comprehensive knowledge", "proven methodologies" feel promotional. Replace with objective language: "Provides ISO 13485 audit guidance including planning, execution, and nonconformity identification."

3. Add concrete validation steps — You mention "Decision Point" twice but never specify how to validate decisions. Add run→check→fix patterns (e.g., "Execute audit-schedule-optimizer.py → validate all processes have risk scores 1-5 → if validation fails, review risk-methodology.md").

4. Standardize terminology — You use "nonconformity" and "finding" interchangeably. Pick one or clearly define the relationship.

Quick Wins

• Most impactful: Create the 3-4 core reference files (especially iso13485-audit-guide.md) — nets you ~7 points
• Next: Add TOC to SKILL.md and strip marketing language — ~3 points
• Then: Add validation/verification steps for the audit workflow — ~2 points
• Polish: Unify terminology and improve Ease of Use triggers — ~1-2 points

Checkout your skill here: [SkillzWave.ai] | [SpillWave] We have an agentic skill installer that install skills in 14+ coding agent platforms. Check out this guide on how to improve your agentic skills.