Contributions are welcome when they make a skill more specific, more useful, or easier to verify.
- a recurring AI tell with a clear reader-facing cost and a before-and-after example
- a detector pattern with a must-fail case and a must-pass case
- a false positive removed from
slop-detector.json, with the sentence that exposed it - a missing edge case, state, or accessibility check in the design skills
- clearer wording that keeps a rule short enough to act on
Avoid rules based only on taste, and avoid declaring a word bad in every context. The severity model exists because most tells are evidence, not proof.
- Edit the relevant
SKILL.md, reference file, orslop-detector.json. - If you touched the detector, add your cases to
slop-detector/scripts/test-severity.py. - Run
./scripts/validate.sh. - Describe the reader-facing problem the change solves.
stop-slop, slop-detector, and the repository scaffolding are MIT. The five impeccable-derived skills (clarify, distill, quieter, polish, harden) are Apache-2.0, and red-pen is redistributed from Rich Tabor without a license grant. Contributions to a folder are licensed under that folder's license. See NOTICE.md.