antipattern-detector

I checked out your antipattern-detector skill and noticed you're tackling the "invisible debt" problem—code that technically works but slowly poisons a codebase. The analysis framework you built (84/100) handles the hard part well, though I'd be curious about your approach to reducing false positives when patterns cluster.

Links:

• GitHub
• Your agentic skill in the SkillzWave marketplace

The TL;DR

You're at 84/100, solid B territory. This evaluation is based on Anthropic's agent skills best practices—specifically the Progressive Disclosure Architecture (PDA), ease of use, spec compliance, and utility. Your strongest area is Utility (18/20) with 25+ patterns across 5 categories that genuinely solve a real problem. The weaker spots are PDA (23/30) and Spec Compliance (12/15), mostly structural issues rather than content problems.

What's Working Well

• Comprehensive pattern library: Your antipattern-catalog.md with 537 lines of concrete examples (scope creep, premature optimization, god objects, etc.) gives developers actual patterns to recognize. This is the meat that makes the skill valuable.
• Clear severity framework: The classification system (critical/high/medium/low) with specific detection signals is actionable. "Look for these phrases" and "Red flags include..." give concrete entry points instead of vague guidance.
• Smart output template: You've structured the detection output as a concrete deliverable—pattern name, severity, symptoms, recommendations—that teams can actually use in code review conversations.
• Solid trigger phrases: The description hits the use case directly ("strategic-cto-mentor needs to identify red flags"), making discoverability straightforward.

The Big One: File Structure & Navigation

Your biggest opportunity is restructuring how the skill organizes reference material. Right now, antipattern-catalog.md lives in the root of the skill folder, but per best practices it should be in a references/ subdirectory—this keeps the main SKILL.md lean while making detailed content accessible on-demand.

The fix: Create antipattern-detector/references/antipattern-catalog.md and update the link in SKILL.md. Also, both files exceed 100 lines but lack tables of contents—add one after each frontmatter so people can jump to specific sections. This alone nets you +3-5 points on PDA by improving token efficiency and navigation.

Bonus issue: You reference ../assumption-challenger/challenge-questions.md in the catalog, but that file doesn't exist in this package. Either remove that reference or include it.

Other Things Worth Fixing

1. Second-person voice - Lines like "you're testing" should be imperative: "test hypothesis" or "validate findings". Keeps the tone consistent and saves tokens (+1 point).

2. Narrative fluff in the intro - Phrases like "Detecting them early saves months of pain" add emotional color but no instruction. Tighten to: "Anti-patterns are proven failure modes that lead to: [list]" (+1 point).

3. Missing validation loop - Your Detection Process has Scan→Verify→Document, but no explicit Step 4 to cross-check findings against historical examples. Add a validation checklist so users don't accidentally flag false positives (+1 point).

Quick Wins

• Move antipattern-catalog.md to references/ folder and add TOCs to both files (+3-5 points)
• Switch second-person voice to imperative tone (+1 point)
• Remove narrative phrases and tighten intro (+1 point)
• Add validation step to detection process (+1 point)

If you nail all four, you're looking at 90-92/100, solid A territory.

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.