Stop Slop
A skill file that teaches Claude (or any LLM) to catch and remove AI writing patterns from prose. By Hardik Pandya, MIT. As of 2026-07-03, GitHub showed 13,255 stars, 928 forks, no public releases/tags, and HEAD
8da1f030185bdfe8471220585162991eaeb970e9. Source: hardikpandya/stop-slop README, GitHub
stop-slop is a single-purpose SKILL.md plus reference files. Its premise: AI writing has predictable phrases, structures, and rhythms, and a skill can teach a model to detect and strip them. It installs as a Claude Code skill, a Claude Projects knowledge upload, custom instructions, or a system-prompt include where reference files load on demand. Source: hardikpandya/stop-slop README, GitHub
The downloaded tweet artifact shows the upstream folder shape: SKILL.md plus references/phrases.md, references/structures.md, and related reference files. That matters because the repository is a full skill package, not just a prompt snippet. Source: X bookmark media artifact 2058483501624668249, reviewed 2026-06-30
What It Catches
- Banned phrases (
references/phrases.md): throat-clearing openers, emphasis crutches, business jargon, adverbs, vague declaratives, meta-commentary. - Structural cliches (
references/structures.md): binary contrasts, negative listings, dramatic fragmentation, rhetorical setups, false agency, narrator-from-a-distance voice, passive voice. - Sentence-level rules: no Wh- sentence starters, no em dashes, no staccato fragmentation, no lazy extremes, active voice required.
It scores prose 1-10 on five dimensions (Directness, Rhythm, Trust, Authenticity, Density); below 35/50 means revise. Source: hardikpandya/stop-slop README, GitHub
Relation to Humanizer
stop-slop is the upstream public analog of Kevin's installed Security and Review Skills skill. Both target the same problem (removing AI tells), and both enforce the no-em-dash rule that runs through Kevin's voice profile. Humanizer is the better fit for Kevin because it carries a permanent voice profile calibrated from his own writing samples and adds spoken-answer calibration; stop-slop is a clean, model-agnostic baseline worth mining for additional banned-phrase coverage. The two converge on Wikipedia's "Signs of AI writing" lineage. Source: X/@tom_doerr, 2026-05-24
Deep-review decision: do not install a second active Kevin writing skill. The Security and Review Skills skill remains the MECE owner for de-AI-writing work. This review folded stop-slop's useful reference checks into skills/personal/humanizer/SKILL.md: false agency, narrator-distance voice, and the five-axis quick score (Directness, Rhythm, Trust, Authenticity, Density). Keep stop-slop as an upstream reference or generic baseline, not a duplicate local route. Source: X bookmark artifact audit, 2026-06-30
The later Guri Singh bookmark is a duplicate source signal, not a new tool. Its X image returned 403 during media repair, but the text names the same single-SKILL.md package and the repo snapshot above remains the current authority. This confirms continued demand without changing the local adoption decision. Source: X/@heygurisingh and GitHub API, 2026-07-03
Timeline
- 2026-07-03 | Reviewed the duplicate Guri Singh source row, noted the X CDN image 403, and refreshed current repo state: 13,255 stars, 928 forks, MIT license, no releases/tags, HEAD
8da1f030185bdfe8471220585162991eaeb970e9. Source: X/@heygurisingh and GitHub, 2026-07-03 - 2026-06-30 | Deep-reviewed upstream repo state: 13,052 stars, 911 forks, MIT license, no releases/tags, HEAD
8da1f030185bdfe8471220585162991eaeb970e9. Folded useful checks into Kevin's Security and Review Skills instead of creating a duplicate skill. Source: GitHubhardikpandya/stop-slop, 2026-06-30 - 2026-05-24 | @tom_doerr surfaced stop-slop ("Removes AI writing patterns from prose"), 1,063 likes, 1,540 bookmarks. Repo at ~7.9K stars (github resolved: 7,937). Source: X/@tom_doerr, 2026-05-24
- 2026-03-18 | Earlier viral signal: @heygurisingh framed stop-slop as "the cheat code for making AI writing undetectable… a single SKILL.md," 867 likes, 1,951 bookmarks — a second independent bookmark for the same repo, confirming sustained traction. Source: X/@heygurisingh, 2026-03-18