Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
110 changes: 34 additions & 76 deletions codex/skills/unslop/SKILL.md
Original file line number Diff line number Diff line change
@@ -1,82 +1,40 @@
---
name: unslop
description: "Writing and editing: remove recognizable AI tells and add human voice. Apply to user-facing prose, docs, explanations, PR text, and other natural-language output."
description: "Remove canned AI-writing patterns from explanations, documentation, PR text, and public prose. Use when a draft is inflated, formulaic, vague, or repetitive, or when the user asks to unslop writing."
---

# Unslop

Edit text to remove AI patterns and add human voice.

## Process

1. Scan for the patterns below.
2. Rewrite. Preserve meaning and match the intended tone.
3. Add soul.
4. Self-audit: "What makes this obviously AI generated?" Fix the remaining tells.

## Adding soul

Removing patterns is half the job. Sterile, voiceless writing is just as obvious.

- **Have opinions.** React to facts instead of neutrally listing pros and cons.
- **Vary rhythm.** Short sentences. Then longer ones that take their time. Mix it up.
- **Acknowledge complexity.** "Impressive but also kind of unsettling" beats "impressive."
- **Use "I" when it fits.** First person isn't unprofessional.
- **Let some mess in.** Perfect structure looks machine-made.
- **Be specific.** Not "this is concerning" but "there's something unsettling about agents churning away at 3am."

## Patterns to detect and fix

### Content

1. **Puffery.** "pivotal moment", "testament to", "evolving landscape", "setting the stage for", "indelible mark", "deeply rooted". Cut puffery, state what happened.
2. **Name-dropping.** Listing media outlets without context. Pick one, say what was said.
3. **Superficial -ing phrases.** "highlighting...", "ensuring...", "reflecting...", "showcasing...", "fostering...". Delete or expand with real sources.
4. **Promotional language.** "nestled", "vibrant", "breathtaking", "groundbreaking", "renowned", "stunning", "must-visit". Use neutral descriptions.
5. **Vague attributions.** "Experts believe", "Industry reports suggest", "Some critics argue". Name the source or delete.
6. **Formulaic challenges.** "Despite challenges... continues to thrive." Replace with specific facts.

### Language

7. **AI vocabulary.** Additionally, crucial, delve, enduring, enhance, fostering, garner, interplay, intricate, landscape (abstract), pivotal, showcase, tapestry (abstract), testament, underscore, vibrant. Replace with plain words.
8. **Fancy ways to say "is".** "serves as", "stands as", "boasts", "features". Just say "is" or "has".
9. **"Not just X, but Y."** State the point directly instead.
10. **Rule of three.** Forcing ideas into groups of three. Use the natural number.
11. **Synonym cycling.** Protagonist, main character, central figure, hero all in one paragraph. Pick one, repeat it.
12. **False ranges.** "from X to Y" where X and Y aren't on a meaningful scale. List topics directly.

### Style

13. **Em dash overuse.** Avoid em dashes entirely. Use periods or commas. If a thought needs separation, end the sentence.
14. **Colon overuse.** Colons are fine before a list or example. Not as mid-sentence connectors. Rewrite so the point stands on its own.
15. **Boldface overuse.** Don't bold every proper noun or acronym.
16. **Inline-header lists.** The tell is a bold label and colon that restates the line: "**Performance:** Performance improved...". Convert those to prose. A bold lead-in that names the item and is followed by genuinely new detail is fine.
17. **Title case headings.** Use sentence case.
18. **Decorative emojis.** Remove them from headings and bullets.
19. **Curly quotes.** Replace with straight quotes.

### Communication artifacts

20. **Chatbot phrases.** "I hope this helps!", "Let me know if...", "Of course!", "Certainly!", "Found the smoking gun!" Remove.
21. **Cutoff disclaimers.** "While specific details are limited..." Find sources or remove.
22. **Sycophantic tone.** "Great question! You're absolutely right!" Respond directly.

### Filler

23. **Filler phrases.** "In order to" becomes "To". "Due to the fact that" becomes "Because". "It is important to note that" gets deleted.
24. **Excessive hedging.** "could potentially possibly be argued that it might" becomes "may".
25. **Generic conclusions.** "The future looks bright." State specific plans or facts.

### Jargon

26. **Abstract metaphor nouns.** Substrate, wedge, vector, locus, vantage, nexus, primitive (as noun), harness (as metaphor), surface (as in "API surface"), bedrock, scaffolding (as metaphor), modality, paradigm, gold-plating, ratchet (as metaphor), evacuate (for moving code), endgame, north star, flywheel. These read as technical but usually have a plainer concrete word. "Substrate" becomes "base". "Wedge in" becomes "add". "Vector" becomes "way" or "method". "Gold-plating" becomes "more than the job needs". "Ratchet" becomes the mechanism's real name or "a limit that only tightens". "Evacuate" becomes "move out". "Endgame" becomes "the last phase". Pick the concrete word.

### Plain speech

27. **Say what it does, not how it feels.** "the database stays close at hand", "SQL you can read", "types that follow your schema" name a feeling. The fix names the mechanism or a number: "`.toSQL()` returns the exact string sent to the database", "a column rename fails the build". Ask what the sentence tells the reader to do or know, then write that. If you can't restate it as a concrete instruction, fact, or number, cut it. If the sentence could appear unchanged in another project's docs, it says nothing about this one. Cut it.
28. **Shorten or split dense sentences.** If the reader has to backtrack to parse a sentence, break it in two or drop clauses. One idea per sentence.
29. **Active voice.** Prefer it. Catch "is/are/was/were + past participle" and name the actor: "queries are validated" becomes "the compiler validates queries". Passive is fine only when the actor is unknown or genuinely doesn't matter.
30. **Cut adverbs, or use a stronger verb.** "runs quickly" becomes "is fast" or the number. "significantly improves" becomes the measured delta. An adverb propping up a weak verb means the verb is wrong.
31. **Prefer the plain word.** "utilize" becomes "use", "leverage" becomes "use", "facilitate" becomes "help", "numerous" becomes "many", "in the event that" becomes "if". The fancier synonym is rarely clearer.

Source provenance lives in [references/sources.md](references/sources.md). Do not load it during normal writing.
Edit the writing to remove AI patterns. Preserve its meaning and intended tone;
do not invent opinions, anecdotes, or personality to make it seem human.

## Patterns to notice

- **Empty praise.** "Groundbreaking", "seamless", and "a testament to" often
stand in for facts. State what happened or how it works. Drop unsupported
claims rather than inventing evidence for them.
- **Vague authority.** "Experts believe" and "industry reports suggest" need
an identifiable source. Name it if supported; otherwise remove the attribution
and check whether the claim itself has support.
- **Inflated language.** Prefer "use" to "utilize" and "is" to "serves as".
Replace decorative metaphors with their literal meaning. Keep technical terms
when they are precise and useful to the intended reader.
- **Mechanical structure.** Forced contrasts, three-part lists, and generic
conclusions give unrelated material the same shape. Use the natural number
of points; keep a contrast when the alternative helps explain the choice.
- **Redundant presentation.** Cut filler such as "it is important to note" and
bold labels that merely repeat the sentence. Use headings, emphasis, and lists
when they help the reader navigate, compare, or follow steps.
- **False variety.** Calling the same thing a tool, platform, solution, and
ecosystem can obscure what it is. Repeat the accurate name. Cut stacked
hedges such as "could potentially possibly" while retaining qualifications
that change the claim.
- **Canned rapport.** "Great question!", automatic agreement, and stock closing
offers can sound pasted on. Answer directly. Warmth, humor, and a distinctive
voice can stay when they belong to the passage.

These are clues to a writing problem, not a banned-word list. Change a passage
where the pattern weakens it; leave useful structure and working sentences alone.

Source provenance lives in [references/sources.md](references/sources.md). Do not
load it during normal writing.
21 changes: 19 additions & 2 deletions codex/skills/unslop/references/sources.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,13 +2,30 @@

Read this provenance when auditing or revising Unslop, not during normal writing.

The skill is substantially derived from Lauren Tan's `unslop` skill in pstack:
Compass originally adapted Lauren Tan's `unslop` skill in pstack:

- Source: https://github.com/cursor/plugins/blob/51a96e0dd838404da19ba83dc70aa21eef71f868/pstack/skills/unslop/SKILL.md
- Repository license: MIT
- Upstream copyright: Copyright (c) 2026 Lauren Tan

Compass keeps the upstream failure catalog and forceful wording where they change model behavior, while pruning or simplifying a small number of house-style details.
The September 5, 2026 revision consolidates the upstream pattern catalog while
removing the instruction to add personality and blanket rules about formatting
or vocabulary. Patterns remain useful editorial clues: lists, contrasts, and
technical terms can be appropriate. Unslop applies to prose with recognizable
AI-writing problems, including technical explanations and PR text. The user's
detailed source-voice preferences belong to their website's editorial guidance.

Sources read September 5, 2026:

- [GPT-6 Astra prompting guidance](https://developers.openai.com/api/docs/guides/latest-model#instruction-following): audit skills because stronger instruction following makes their contents more consequential.
- [Fable 5 prompting guidance](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-fable-5#recommended-scaffolding-changes): older, prescriptive skills can reduce quality; compare against defaults.
- [Fable 5.1 prompting guidance](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-fable-5-1#formatting-in-chat): older anti-formatting instructions can now suppress useful structure.
- [Matt Pocock's authoring guidance](https://github.com/mattpocock/skills/blob/3cca18b368ae95cdbdebbff572ccafa662551015/skills/productivity/writing-for-agents/SKILL.md): delete instructions that do not change behavior relative to the current model. This is practitioner guidance, not a GPT-6 comparison.

These sources support revisiting instructions and retaining useful distinctions.
They do not establish that every older rule is redundant or that this particular
catalog improves every draft. The accompanying comparisons are limited editing
checks, not a general quality benchmark.

## Upstream license

Expand Down