slop
Detect AI-generated design slop with concrete, locatable findings. Use when the user wants to check code for AI authorship tells, design inconsistency, or generic UI patterns. Each finding includes fi
SKILL.mdskills/claudskills/slop/SKILL.md
GitHub/slop
The viral hook of dragoon. Detects the patterns that betray AI-generated UI: 7 different shadow values, generic violet-to-pink gradients, transition-all-duration-200 everywhere, lorem ipsum left in source, off-grid spacing, off-palette colors.
When to use
Run /slop when:
- The user wants to check if code "looks AI-generated" or "feels off"
- Reviewing a PR or diff before merging
- After a feature lands, to catch slop before it ships
- The user says "find the slop", "scan for AI tells", "audit for design hygiene"
How it works
12 deterministic rules. Each rule:
- Returns concrete findings with line, column, snippet, message, and fix
- Has a severity (low / medium / high)
- Is unit-tested
Two rules become codebase-aware when dragoon.json is present:
slop-006spacing-off-grid uses your detected gridslop-007hardcoded-color-off-palette uses your detected palette
So /slop gives generic results without /scan, and project-specific results with it. Always run /scan first if you can.
Run it
node ~/.claude/skills/dragoon/skills/slop/scripts/slop.js [path]
Flags:
--manifest <p>use a specific dragoon.json (auto-detects ./dragoon.json otherwise)--jsonmachine-readable--severity <low|medium|high>filter--ruleslist all rules and exit--quietonly show findings, no summary--helpfull usage
Exit codes
0no findings1findings present (good for CI gating)2bad usage
All rules
Run node ~/.claude/skills/dragoon/skills/slop/scripts/slop.js --rules for the full list with descriptions.