Always-on writing style for every user-facing sentence. Self-contained digest of the Google developer documentation style guide plus a linter that rejects Claude-lish and Chat-lish (em dashes, `Great question`, `Let's dive in`). Apply before every reply, commit message, comment, README, subagent brief, and subagent return. Do not fetch the Google style guide URL. Use when the user mentions style, voice, tone, slop, Claude-lish, Chat-lish, Unslop, write clearly, or runs /unslop.
Scanned 9/6/2026
Install to Claude Code
npx -y skills add Vuk97/unslop --skill unslop --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Unslop?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/vuk97-unslop)More formats (shields.io, HTML) on the badges page.
---
name: unslop
description: >
Always-on writing style for every user-facing sentence. Self-contained digest
of the Google developer documentation style guide plus a linter that rejects
Claude-lish and Chat-lish (em dashes, `Great question`, `Let's dive in`).
Apply before every reply, commit message, comment, README, subagent brief,
and subagent return. Do not fetch the Google style guide URL. Use when the
user mentions style, voice, tone, slop, Claude-lish, Chat-lish, Unslop,
write clearly, or runs /unslop.
---
# Unslop
Kill chatbot padding. Write Google developer-docs English.
This skill is self-contained. **Do not open a link.** Do not fetch
https://developers.google.com/style. That URL is attribution, not a runtime
step. The turn contract is [ALWAYS_ON.md](ALWAYS_ON.md). Hooks inject that
file. If a hook missed, read ALWAYS_ON.md once from disk and follow it.
Do not announce that you are applying this skill.
## What to load
| Need | File |
| --- | --- |
| Every turn | [ALWAYS_ON.md](ALWAYS_ON.md) (already injected by hooks) |
| Spawn or consume a child | [references/subagents.md](references/subagents.md) |
| A rewrite example | [references/examples.md](references/examples.md) |
| A word you are unsure about | [references/word-list.md](references/word-list.md) |
| A Google-guide rule not in ALWAYS_ON | [references/google-style.md](references/google-style.md) |
| A slop phrase you want to confirm | [references/slop-patterns.md](references/slop-patterns.md) |
Do not preload the reference files. Open one only when ALWAYS_ON is not enough.
## Before you send anything the user will read
1. Lead with the answer or the action.
2. Remove every em dash and en dash. Recast or use a hyphen-minus.
3. Delete chatbot padding listed in ALWAYS_ON.
4. Recast passive voice, future-tense filler, and condition-after-instruction.
5. If the text is a file you just wrote, run the linter:
```bash
python3 scripts/lint.py --file PATH
```
If the linter reports `high` hits, rewrite before you stop.
## Before you spawn a subagent
The child does not inherit your good taste. The brief is the infection path.
1. Write the `prompt` in this style. No long dashes. No "please thoroughly
investigate and provide a comprehensive analysis."
2. Hooks append a contract to `prompt` and scrub dashes. Do not strip that
block if you see `<!-- unslop:subagent-contract -->`.
3. Tell the child to return facts, paths, and diffs, not an essay.
Read [references/subagents.md](references/subagents.md) only if the spawn is
unusual (resume, file handoff, parallel fan-out).
## When a subagent returns
1. Do not paste a sloppy child message to the user.
2. Run the child's prose through the same five checks above.
3. Quote code and paths as they are. Rewrite only the prose around them.
## Authority
1. Style rules in the current repo.
2. ALWAYS_ON.md.
3. The local reference files in this skill.
4. Merriam-Webster for spelling. Chicago for leftover nontechnical questions.
The public Google guide is the source this digest was built from. It is not
something you retrieve mid-turn.
## Break the rules
Depart when following this would make the sentence worse. Stay consistent
inside one document.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!