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.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-006 spacing-off-grid uses your detected grid
  • slop-007 hardcoded-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)
  • --json machine-readable
  • --severity <low|medium|high> filter
  • --rules list all rules and exit
  • --quiet only show findings, no summary
  • --help full usage

Exit codes

  • 0 no findings
  • 1 findings present (good for CI gating)
  • 2 bad usage

All rules

Run node ~/.claude/skills/dragoon/skills/slop/scripts/slop.js --rules for the full list with descriptions.