detect-ai-sloplisted

Use when the user wants to SEE which AI tells are in a piece of text without rewriting it. Triggers include "flag the AI slop", "where are the tells", "show me what to fix", "detect AI-generated writing", "check this for slop but don't change it", "what would you flag here", or any editorial / pre-publish review that lists issues by location. This is the report lens; it returns located findings (category, severity, the offending text), never a rewrite and never a single 0-10 score. To rewrite instead, use remove-ai-slop.
SalZaki/antislop · ★ 0 · AI & Automation · score 72

Install: claude install-skill SalZaki/antislop

# Detect AI Slop Report where the AI tells are. Return located findings, grouped by category, so the author knows exactly what to fix. This lens never rewrites and never emits a single headline score — a number invites gaming, a located finding invites editing. Rules, categories, severity, and the override-merge order are defined once in [`../../shared/spec.md`](../../shared/spec.md). Read it before reporting. The countable tells are scored by a deterministic script; the meaning-dependent ones are yours to judge. ## Two tiers **Deterministic tier (preferred).** Run the bundled script. It counts the countable tells (vocabulary-table hits, em-dash density, fixed templates, bold/bullet soup, fixed scaffolding phrases) and returns stable JSON. ```bash python3 shared/slop_count.py --file <path> \ [--allow-list ~/.claude/config/remove-ai-slop/user-allow-list.md] \ [--allow-list .claude/skills/remove-ai-slop/overrides/user-allow-list.md] \ [--extra-vocab ~/.claude/config/remove-ai-slop/user-vocabulary.md] \ [--extra-vocab .claude/skills/remove-ai-slop/overrides/user-vocabulary.md] ``` Pipe text on stdin instead of `--file` when the user pasted it. Pass whichever override files exist (skip silently if absent). The script is the **sole parser** of the override files: consume its `findings` and `summary`, do not re-parse the override tables yourself. **Judged tier (you).** The script cannot see meaning. After the script runs, add findings for the meaning-dependent catego