Skip to content

petems/unsloppify

Repository files navigation

unsloppify

Strip AI slop from your prose. A fast CLI prose linter for markdown, plus a Claude Code skill and output-style for the judgment calls a regex can't make.

uvx unsloppify draft.md            # report findings, exit 1 if any
uvx unsloppify draft.md --fix      # rewrite safe deterministic fixes in place
uvx unsloppify docs/ --format json # recursive scan, JSON output

Why this exists

Existing Claude skills like stop-slop and anti-ai-slop-writing ask Claude to scan your prose for tells. That works, but it costs tokens and Claude can miss obvious cases.

unsloppify splits the work:

Tier Examples Handled by
1. Punctuation em dashes, smart quotes, ellipsis chars CLI, auto-fixed
2. Banned phrases delve into, tapestry, In today's fast-paced world CLI, flagged or auto-fixed
3. Structures binary contrasts, false agency, parataxis Skill (LLM judgment)

Run the CLI first to nuke the obvious stuff. Let the skill handle the rest.

Install

uvx unsloppify --help               # one-shot, no install
pipx install unsloppify             # persistent CLI
uv add --dev unsloppify             # as a dev dep in a uv project

Usage

CLI

unsloppify FILE [FILE...]           # lint one or more files / directories
  --fix                             # apply safe fixes in place
  --format {text,json,github}       # output format (default: text)
  --severity {error,warning,info}   # minimum severity to report
  --rules PATH                      # add custom rules YAML
  --list-rules                      # print bundled rule catalogue

Exit codes: 0 clean, 1 findings present, 2 invocation error.

Pre-commit

repos:
  - repo: https://github.com/petems/unsloppify
    rev: v0.1.0
    hooks:
      - id: unsloppify

GitHub Actions

- run: uvx unsloppify docs/ --format github

Claude Code skill

Two ways to install:

Plugin marketplace (recommended for Claude Code):

/plugin marketplace add petems/unsloppify
/plugin install unsloppify@unsloppify

Skills CLI (works with Claude Code, Codex, Cursor, and other agents):

npx skills add petems/unsloppify

Then ask the agent to "unsloppify this draft" and it will invoke the CLI first for the deterministic catches, then do a judgment pass on structures.

Claude Code output style

cp output-styles/unsloppified.md ~/.claude/output-styles/ then /output-style unsloppified. Claude's own responses will avoid em dashes, banned phrases, and structural cliches.

Rule catalogue

unsloppify --list-rules

Or browse src/unsloppify/rules/*.yaml directly. Rules are versioned with the package.

Credit

The vocabulary lists and structural patterns are adapted from MIT-licensed upstream skills. See THIRD_PARTY_NOTICES.md.

License

MIT

About

Strip AI slop from prose. Fast Python CLI for markdown + Claude Code skill + output style. Hard-locks em dashes, banned phrases, and AI tells.

Topics

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors