Slopless

Use this skill when asked to review English Markdown prose for AI slop, human slop, weak phrasing, cliches, filler, or formulaic prose.

Scope

• Slopless is English-only.
• Slopless emits JSON only.
• Slopless reports findings. It does not rewrite prose.
• Do not use Slopless as a fact checker.

Required Workflow

1. Run npx slopless --help before the first Slopless run in the session.
2. Create .slopless/findings in the current working directory.
3. Run Slopless on the requested Markdown files, folders, globs, or stdin.
4. Save raw JSON output under .slopless/findings/.
5. Use a timestamped filename that identifies the input.
6. If the user asks for explanation, summarize the saved JSON findings for the user.
7. Do not leave the only useful result in a temp directory.

Commands

mkdir -p .slopless/findings
npx slopless "docs/**/*.md" > ".slopless/findings/$(date +%Y-%m-%d-%H%M%S)--docs-review.json"

mkdir -p .slopless/findings
npx slopless draft.md > ".slopless/findings/$(date +%Y-%m-%d-%H%M%S)--draft.json"

mkdir -p .slopless/findings
npx slopless --stdin --stdin-filename draft.md > ".slopless/findings/$(date +%Y-%m-%d-%H%M%S)--stdin-draft.json"

Output Handling

• Exit 0: no findings.
• Exit 1: findings were reported.
• Exit 2: command failed before linting.
• Treat exit 1 as successful execution with prose findings.
• Read the JSON before explaining results.
• Preserve rule IDs, file paths, line numbers, and excerpts when summarizing.

Ignore Rules

Use textlint comments when a finding is intentionally ignored:

<!-- textlint-disable slopless/semantic-thinness -->

Something shifted in the room.

<!-- textlint-enable slopless/semantic-thinness -->