deslop
Skill36 starsChanged 22 days agoMITView on GitHub
Use this skill when editing prose for a human reader or removing formulaic AI writing patterns. Activates on mentions of deslop, unslop, humanize, remove AI tells, AI slop, sounds like AI, reads like ChatGPT, em dashes, clean up this prose, make this sound human, or slop check. PR bodies and replies use super-good-pr for structure; interactive chat follows the project register.
118 other projects use a version of this.
Adopted by
Most recently taken up by
- lalulali/vespyrtheir own version, first seen Nov 8, 2026
- ferologics/pi-skillstheir own version, first seen Nov 8, 2026
- novuhq/novutheir own version, first seen Nov 8, 2026
- openclaw/openclaw-enterprisetheir own version, first seen Nov 8, 2026
- ADWilkinson/astacktheir own version, first seen Nov 8, 2026
- MammadovEshgin/clean-code-skilltheir own version, first seen Nov 8, 2026
Used with
Projects with it also use simple-issue-description, control-cli and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
JwuthriSkillsimple-issue-description
Turn a rough bug report, feature request, support note, or pull request into a short, plain-language issue focused on the problem and desired behavior. Use when a contributor asks to simplify an issue, explain what a PR is for, create the corresponding issue for a PR, remove implementation detail from a report, or invokes /simple-issue-description.
Also in 4 projects that use this
ElyzeSolutionsSkillcontrol-cli
Build or adapt a local harness to drive, inspect, and profile an interactive CLI or TUI without external services. Use for CLI UX checks, startup regressions, memory leaks, hangs, prompt flows, or terminal demos.
Also in 6 projects that use this
JwuthriSkillopenseo-review-web-content
Write and review content for the OpenSEO website (web/) — blog posts, guides, feature pages, FAQs. Distills the philosophy for on-brand, useful, accurate content. Use whenever adding or editing user-facing prose in web/content or web/src.
Also in 4 projects that use this
learnerLjSkillverify-this
Verify a claim with fresh local evidence: restate it falsifiably, capture baseline and treatment, compare artifacts, and return VERIFIED, NOT VERIFIED, or INCONCLUSIVE.
Also in 6 projects that use this
ElyzeSolutionsSkillcontrol-ui
Build or adapt a local browser/CDP harness to drive and inspect a web, IDE, or Electron UI. Use for local UI verification, screenshots, accessibility snapshots, perf profiles, visual diffs, or reproducing UI bugs.
Also in 7 projects that use this
ten9876Skillpapercuts
Read-only AetherSDR issue triage that surfaces the UX papercuts, bugs, and regressions you can fix AND prove today with the agent automation bridge. Use when the user says $papercuts or /papercuts, or asks to "find papercuts", "what can I fix and prove today", "triage bridge-provable issues", "what UX bugs can the bridge close", "sort issues by priority I can knock out", or otherwise wants a curated, lifecycle-aware (design→code→prove) shortlist of issues — NOT the label-and-respond /triage workflow. This is the discovery step that feeds /fb. Strictly read-only: it never labels, comments, closes, or edits issues.
Also in 4 projects that use this
- +5 more from Jwuthri
Other versions118 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 118
- molt-bot/clawdbot★ 391ksame name
- novuhq/novu★ 40ksame name
- every-app/open-seo★ 27ksame name
- millionco/react-doctor★ 15ksame name
- jaredpalmer/kev★ 9.8ksame name
- joe-bell/cva★ 6.9ksame name
- sanity-io/sanity★ 6.4ksame name
- millionco/expect★ 3.6ksame name
- rohitg00/pro-workflow★ 2.9ksame name
- compozy/compozy★ 2.9ksame name
108 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository14 other files it gives its agents.
- AGENTS.mdAGENTS.md
- brainstormSkill
- codex-imagegenSkill
- cross-model-reviewSkill
- demo-reelSkill
- dreamSkill
- gitSkill
- hyper-pr-reviewSkill
- implementSkill
- orchestrateSkill
- planSkill
- researchSkill
- super-good-prSkill
- tui-designSkill
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 7 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 116 edited versions we read keep these.
7
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Remove formulaic writing that obscures meaning or flattens the author's voice.
- Distinctive: line 10
Preserve meaning first: Keep facts, qualifications, attribution, and the author's stance.
- Distinctive: line 54
Run the bundled scanner from the repository root, or resolve its path relative to this skill's directory when installed elsewhere:
- Very specific: line 67
Treat candidates as prompts to inspect, never automatic replacements.
- Distinctive: line 71
Start with missing meaning or a confusing argument, then work toward sentence detail.
- Distinctive: line 87
Keep every substantive fact unless the user requested substantive cuts.
- Distinctive: line 89
Keep protected legal text and meaningful safety instructions even when they sound formulaic.
- Distinctive: line 112
Default to embedded when another skill invokes this one.
“Common” is measured: kept by at least 6 in 10 of the 116 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Deslop
Remove formulaic writing that obscures meaning or flattens the author's voice. Name the reader's problem before editing: an unsupported claim, a repeated conclusion, an unclear actor, or a sentence that needs rereading. A recognizable word or construction alone is not a defect.
Preserve meaning first. Keep facts, qualifications, attribution, and the author's stance. Improve clarity within the requested scope. Never invent measurements, mechanisms, anecdotes, or confidence to make a vague sentence sound specific. NO_EDITS is a successful result when the draft already works.
Surface first
The user, publication, and project contract set the register. An artifact's authority skill decides its structure; this skill supplies prose diagnostics.
| Surface | Authority and preservation |
| PR body, PR comment reply | super-good-pr; preserve semantic emoji, reviewer guidance, receipts, and human-authored choices |
| PR review report | hyper-pr-review; preserve severity and finding structure |
| Commit body | Repository commit rules; plain text and required trailers |
| Spec, plan, research report | Preserve decisions, alternatives that matter, uncertainty, citations, and requirements |
| README, docs, guides | Explain the actual system; preserve commands, examples, signatures, and error strings |
| Changelog, release notes | Narrating the change is the purpose; keep versions and migration consequences |
| Slack, email | Match the sender's own messages, not the channel's; preserve signatures and thread links |
| Blog, essay, fiction | Match the author; creative invention is allowed only within the creative brief |
| Scientific, legal, medical, postmortem | Preserve calibrated uncertainty and approved language |
| Marketing | Follow the brief; persuasion does not authorize invented claims |
| Agent brief, skill, memory | Keep useful tables and terminology; clarity and factual checks still apply |
| Interactive chat | Skip a separate pass; follow the project register |
Read references/surface-profiles.md when a surface needs a worked example. A writing sample guides voice only when its author is the one you are writing for; other people's and other agents' messages in a thread are context, not a sample. A sample does not override an explicit instruction from the user.
A focused editing loop
Use the depth the artifact needs. A short message can be edited in one pass. For a long or consequential document, separate diagnosis from rewriting so the rewrite has an explicit target. Independent review can help when semantic drift would be costly; a second model is not proof of correctness.
digraph deslop {
rankdir=LR;
node [shape=box];
scope [label="Identify reader and protected content"];
diagnose [label="Name a concrete defect"];
edit [label="Make the smallest useful edit"];
verify [label="Compare meaning and scan"];
scope -> diagnose -> edit -> verify;
verify -> diagnose [label="specific unresolved defect"];
}For an audit, report the span, defect, and suggested edit without changing the artifact. For an authorized rewrite, keep the diagnosis internal unless the user asks for it. Do not manufacture a finding quota or require several weak markers before fixing one clear ambiguity.
Mechanical scan
Run the bundled scanner from the repository root, or resolve its path relative to this skill's directory when installed elsewhere:
perl skills/deslop/scripts/slopscan.pl --surface house FILE...
# Surfaces: house | published | agent | sample
# --show-masked lists protected hits suppressed by the masker.
# Exit: 0 no findings, 10 candidates, 20 hard findings, 30 unreadable input.Choose house for our prose, agent for our instruction files, and published or sample for another author's work. The house dash ban is a style requirement, not evidence of authorship. Typography on other surfaces needs contextual judgment (a non-breaking space may be intentional localization).
The scanner is read-only and uses a heuristic Markdown masker. It protects frontmatter, code, blockquotes, comments, and link targets. It also reports raw hard hits suppressed by masking; inspect those when quoting or nested lists make the mask uncertain. A known limitation is a list fence whose closing indentation exceeds three spaces. A clean exit establishes only that the enabled checks found nothing in the visible text they recognized. It does not establish prose quality or parser completeness.
Treat candidates as prompts to inspect, never automatic replacements. Rhythm diagnostics are descriptive: short technical instructions can legitimately have uniform sentence lengths. Run scripts/selftest.sh after changing scanner behavior.
Read in order of consequence
Start with missing meaning or a confusing argument, then work toward sentence detail. These are diagnostic lenses, not five mandatory invocations:
| Lens | Look for | Useful move |
| Document | Repeated summaries, decorative headings, reasoning split into unrelated bullets | Keep the structure that helps the reader navigate or act |
| Argument | Inflated significance, unsupported attribution, imaginary counterarguments | State the actual claim and its support |
| Sentence | Unclear actor, buried action, clause stacks, empty participial tails | Give the relevant subject a clear verb; split only where meaning benefits |
| Vocabulary | Filler, imprecise metaphor, jargon the reader cannot interpret | Use the precise familiar term; preserve defined terminology |
| Evidence | Invented specificity, overstated certainty, stale claims, process narration | Restore qualifications and attribution; flag source gaps |
The catalog in references/pattern-catalog.md gives stable IDs and examples. Consult relevant entries instead of loading every rule for every artifact. The craft moves in references/craft-moves.md help when a sentence's shape is the problem.
Preservation contract
- Preserve code, commands, configuration, quoted language, links, and required metadata unless their modification is part of the request.
- Preserve negation, quantities, units, scope, time, and causal strength. Changing "may" to "will" or "associated with" to "causes" changes the claim.
- Keep every substantive fact unless the user requested substantive cuts. Remove repetition only when its function is also redundant (reminders in procedures can be necessary).
- Reuse defined terms. Synonym variation can accidentally invent a second component or loosen a contractual term.
- Keep protected legal text and meaningful safety instructions even when they sound formulaic.
- Match the author's voice. Do not manufacture errors, arbitrary sentence variation, personal experience, or artificial informality.
Examples in the references are illustrative. Added specifics are usable only when the source material supports them. If a vague draft lacks the necessary facts, keep a plain version or flag what evidence is missing.
Verify the edit
Compare the original and revised text for changed meaning before scanning typography. Check that every new factual clause has support and that deleted clauses did not carry a requirement, exception, or qualification. Use the actual diff for files; reread the complete artifact when edits changed its structure.
Fix identified regressions and repeat the relevant check. Stop when the requested improvements are complete and no concrete defect remains. Do not chase a detector score, a sentence-length target, a percentage reduction, or a vague feeling that a text still sounds generated.
For difficult passages, useful questions are: what would deletion lose; what does a named source actually support; could this paragraph describe any project; and does the same term mean the same thing throughout? A required boilerplate paragraph can pass even when it is portable.
Invocation modes
| Mode | Result |
| Embedded | Return or insert the finished text without an audit preamble |
| File | Edit the authorized prose and summarize material changes |
| Pasted | Return the revision; explain the edits when requested |
| Gate | Report concrete unresolved defects and applicable style violations |
| Audit | Give located feedback without rewriting or judging authorship |
Default to embedded when another skill invokes this one. An audit request does not authorize file edits or publication.
Research boundary
Studies show differences between particular model outputs and human corpora. They do not establish universal forbidden constructions or predict Astra's writing on a new task. Technical prose, fiction, and biomedical abstracts have different requirements. Use research to challenge folklore and identify candidates, then judge the actual text. Verified sources, limits, and maintenance guidance live in references/evidence.md.
Anti-patterns
| Mistake | Correction |
| Fix a wordlist and stop | Check the argument and meaning first |
| Require several markers for an obvious defect | Report the defect on its own evidence |
| Turn a vague claim into an invented measurement | Use supplied evidence or retain a plain limited claim |
| Strip every hedge, passive, heading, or list | Judge its function and the surface |
| Apply our typography to another author's sample | Follow the requested register and explicit constraints |
| Treat scanner output as a rewrite command | Inspect candidates and protected regions |
| Keep rewriting toward an arbitrary score | Stop on satisfied requirements and preserved meaning |
What This Skill is NOT
- An authorship detector or a detector-evasion tool. Never infer who wrote a passage from stylistic markers.
- A mandatory rewrite into one house voice.
- A replacement for an artifact's template, structure authority, or source evidence.
- A full fact-check unless requested. Surface a concrete factual problem encountered during editing instead of silently polishing it into authority.
References
references/pattern-catalog.md: stable diagnostic IDs and examples.references/craft-moves.md: sentence and paragraph repairs.references/surface-profiles.md: artifact-specific preservation and worked rewrites.references/evidence.md: checked research and its limits.scripts/slopscan.pl,scripts/selftest.sh, andreferences/fixtures/: mechanical checks and regression coverage.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

