agentleFS
abduallahJamalabduallahJamal/humanizer-skills/.claude/skills/anti-slop-coder/SKILL.md

anti-slop-coder

Skill0 starsChanged 10 days agoMITView on GitHub

Prevent and surgically clean AI tells in developer artifacts: READMEs, technical docs, PR descriptions, git commit messages, issue replies, and code comments (EN & AR). Blocks narrating comments, AI co-author trailers, em dashes, bold-label bloat, and verbose commit marketing (منع بصمات الذكاء الاصطناعي في الكود والتوثيق والـ README ورسائل Git والتعليقات البرمجية).

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 2 are distinctive, and 11 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 11

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Keep repositories, documentation, pull requests, commit history, and source comments reading like a sharp human engineer wrote them.

    line 12
  2. Very specific:

    Read references/dev-antipatterns.md for side-by-side before/after patterns in English and Arabic.

    line 19
  3. Very specific:

    Delete section-banner ASCII art (// ==================== HELPERS ====================) unless already established across the file.

    line 25
  4. Very specific:

    Write concise subject lines (< 55 chars) naming the concrete change (fix token refresh race on 401, add Arabic RTL support to table header).

    line 29
  5. Very specific:

    Follow the repo's existing commit style (git log -n 10 --oneline) — only add feat: / fix: prefixes if the repo already uses Conventional …

    line 30
  6. Very specific:

    Never: write commit bodies like "This commit introduces a comprehensive and robust module that leverages...".

    line 31
  7. Very specific:

    Do not append Co-authored-by: Claude..., Co-authored-by: Cursor..., Co-authored-by: GitHub Copilot..., or Generated with...

    line 34
  8. Distinctive:

    Follow immediately with copy-pasteable install and usage commands.

    line 38
  9. Very specific:

    Do not format every bullet as - Fast: Lightning-quick processing / - Reliable: Battle-tested.

    line 42
  10. Very specific:

    Avoid prose em dashes (— / -- ).

    line 43
  11. Very specific:

    Avoid awkward calques in Arabic docs: write يعالج الخادم الطلبات instead of يقوم الخادم بمعالجة الطلبات, and أضفنا المصادقة عبر JWT or يدعم …

    line 53
  12. Very specific:

    Keep technical identifiers (JWT, OAuth2, SKILL.md, CLI commands) in backticks without mangling them into ambiguous translations.

    line 54
  13. Very specific:

    When python3 is available, scan target docs or source files before and after cleanup:

    line 60

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Anti-Slop Coder — Developer, Docs, Git & Code Hygiene (EN & AR)

Keep repositories, documentation, pull requests, commit history, and source comments reading like a sharp human engineer wrote them.

Two Operating Modes

  1. Proactive Generation Mode: Activate automatically when drafting READMEs, changelogs, PR descriptions, issue/review replies, architecture docs, commit messages, or code comments.
  2. Surgical Edit-in-Place Mode (edit): Activate when asked to clean AI slop from existing .md / .mdx / .rst docs or code comments in a repo. Make minimal, targeted edits; preserve passages that already read fine, and never touch executable logic, fenced code blocks, CLI flags, URLs, file paths, tables, or YAML frontmatter.

Read references/dev-antipatterns.md for side-by-side before/after patterns in English and Arabic.

Ten Hard Rules for Engineering Artifacts

1. Code Comments: Explain WHY, Never WHAT

  • Ban narrating comments that restate the next line (// Import dependencies, // Loop through users, # استيراد المكتبات, # إرجاع النتيجة).
  • Delete section-banner ASCII art (// ==================== HELPERS ====================) unless already established across the file.
  • Comment only non-obvious invariants, performance traps, browser/OS quirks, protocol specs, or issue links.

2. Git Commits: Terse & Specific, Zero Marketing

  • Write concise subject lines (< 55 chars) naming the concrete change (fix token refresh race on 401, add Arabic RTL support to table header).
  • Follow the repo's existing commit style (git log -n 10 --oneline) — only add feat: / fix: prefixes if the repo already uses Conventional Commits.
  • Never write commit bodies like "This commit introduces a comprehensive and robust module that leverages...". Add a short body only when the why or migration impact isn't obvious from the diff.

3. Never Add AI Git Trailers Unless Asked

  • Do not append Co-authored-by: Claude..., Co-authored-by: Cursor..., Co-authored-by: GitHub Copilot..., or Generated with... trailers unless the user or repository CONTRIBUTING.md explicitly requires them.

4. READMEs & Technical Docs: Substance First

  • Open with one plain sentence stating what the package/tool actually does — never "In today's fast-paced cloud-native landscape..." or "أداة ثورية شاملة لإطلاق العنان...".
  • Follow immediately with copy-pasteable install and usage commands.
  • Cut Happy coding!, Let's dive in!, decorative heading emojis (## 🚀 Features), Title Case on every subheading, and heading-echo first sentences (## Caching\n\nCaching is crucial for performance.).

5. Kill Bold-Label List Disease & Em Dashes in Prose

  • Do not format every bullet as - **Fast:** Lightning-quick processing / - **Reliable:** Battle-tested. Write uneven, concrete bullets or plain prose.
  • Avoid prose em dashes (— / --); use periods, commas, colons, or parentheses. Never touch hyphens/dashes inside code, CLI flags (--profile), or paths.

6. PR Descriptions & Code Review Replies: Lead With the Decision

  • PRs: State (1) what changed, (2) why / linked issue, and (3) how it was verified. Cut "This PR introduces a seamless...".
  • Review / Issue / Slack replies: The reviewer already knows the thread context. Put the decision or answer in sentence 1, followed only by the single constraint or trade-off they haven't seen yet.

7. Do Not Narrate Deleted History Inside Permanent Docs

  • In docs and code comments, describe how the system works now (Uses an LRU cache with 5m TTL), not how you edited it (This function was updated to replace the old O(n^2) loop). Save history for commit messages and CHANGELOG.md.

8. Clean Bilingual Arabic Technical Writing

  • Avoid awkward calques in Arabic docs: write يعالج الخادم الطلبات instead of يقوم الخادم بمعالجة الطلبات, and أضفنا المصادقة عبر JWT or يدعم النظام... instead of repetitive تم إضافة نظام مصادقة شامل يقوم بتوفير....
  • Keep technical identifiers (JWT, OAuth2, SKILL.md, CLI commands) in backticks without mangling them into ambiguous translations.

9. Protect Structured Content During Cleanup

  • Refuse prose-humanization passes on raw source code, JSON/YAML configs, lockfiles, SQL migrations, or test fixtures unless specifically asked to clean comments/docstrings only.

10. Verify With the Repo Slop Scanner

When python3 is available, scan target docs or source files before and after cleanup:

python3 <skill-dir>/scripts/repo_slop_scan.py README.md docs/ src/

Fix every reported prose tell, narrating comment, or trailer hit before finishing.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.