agentleFS
TylerAkinsTylerAkins/ai-agent-skills/skills/anti-ai-writing/SKILL.md

anti-ai-writing

Skill1 starsChanged 34 days agoMITView on GitHub

Draft and check concise technical prose for chat, reviews, replies, tickets, PRs, commits, documentation, and code comments. Use when writing or revising human-facing text; preserve destination contracts and publication approval.

10 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

12 rules. 0 are common, 12 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 8 edited versions we read keep these.

  • 12

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use links supported by the destination: Cursor citations where available, absolute file links in Codex, and Markdown links for GitHub/Jira.

    line 29
  2. Distinctive:

    Remove filler, corporate pleasantries, performative agreement, engagement bait, unnecessary praise, repeated summaries, redundant headings, …

    line 36
  3. Distinctive:

    Verify factual claims, status, and cited locations.

    line 39
  4. Distinctive:

    Read the exact final text before showing it and again after any revision.

    line 44
  5. Distinctive:

    Delete sentences that add no needed fact, evidence, decision, or context.

    line 47
  6. Distinctive:

    Never use em dash punctuation in authored prose.

    line 52
  7. Distinctive:

    Use ordinary punctuation. Avoid decorative arrows, Unicode bullets, ornamental separators, emoji, repeated exclamation marks, and ellipses …

    line 56
  8. Distinctive:

    Prefer direct sentences over rhetorical questions, sentence fragments used for emphasis, and formulaic contrasts such as "This isn't about …

    line 64
  9. Distinctive:

    Remove stock openers and closers such as "Great question," "Absolutely," "Let's dive in," "It's worth noting," "In today's fast-paced …

    line 66
  10. Distinctive:

    Avoid bold labels on every bullet, one-sentence sections, repetitive three-part lists, unnecessary title-case headings, and a closing …

    line 72
  11. Distinctive:

    Before posting an external review, inline comment, thread reply, or status comment, show the exact text and destination and obtain user …

    line 83
  12. Distinctive:

    If the destination or required evidence is missing, label the text as an example or incomplete draft and exclude it from the approval …

    line 90

“Common” is measured: kept by at least 6 in 10 of the 8 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.

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

Technical Writing

Establish Destination and Facts

  1. Identify the reader, destination, requested action, verified facts, and material uncertainty. Preserve intent and unrelated existing content.
  2. Apply the destination's required fields and output contract before generic writing preferences. Preserve quoted evidence, identifiers, schema keys, and machine payloads rather than rewriting them for tone.

Draft Proportionally

  1. Lead with the result, defect, or decision. Use plain language, complete sentences, and concrete evidence. Include only what the reader needs to act or assess it.
    • For a routine clean review, use the reviewing workflow's required body. If none is specified, give a brief, evidence-supported result and state material limits.
    • For a thread reply, usually use 1–3 sentences about that issue's fix, evidence, or question. Retain necessary complexity and uncertainty.
    • For a code comment, explain non-obvious reasons, contracts, precision, or workarounds. Preserve required documentation and useful type annotations.
  2. Use links supported by the destination: Cursor citations where available, absolute file links in Codex, and Markdown links for GitHub/Jira. In other environments, use the renderer's supported link format. Fence suggested code with a language tag.

Edit the Exact Draft

  1. Remove filler, corporate pleasantries, performative agreement, engagement bait, unnecessary praise, repeated summaries, redundant headings, generic reminders, and investigation narration used only to demonstrate effort.
  2. Verify factual claims, status, and cited locations. Distinguish planned, locally changed, checked, pushed, and published states. Claim completed work only when observed, and acknowledge meaningful mistakes plainly. During final diff inspection, remove comments and headers that repeat nearby code or names; retain useful reasons and required documentation.
  3. Read the exact final text before showing it and again after any revision. Apply the strict style rules below to the exact output, including headings, bullets, comments, and publication drafts. Delete sentences that add no needed fact, evidence, decision, or context. Retain blockers and verification limits in the destination allowed by the workflow.

Style rules

  • Never use em dash punctuation in authored prose. Rewrite with a period, comma, colon, or parentheses. Do not substitute double hyphens or spaced en dashes to recreate the same construction. Use ordinary hyphens for compound words and minus signs or range notation where technically appropriate.
  • Use ordinary punctuation. Avoid decorative arrows, Unicode bullets, ornamental separators, emoji, repeated exclamation marks, and ellipses used for dramatic effect. Use standard Markdown lists when a list helps.
  • For GitHub PR descriptions, review bodies, inline comments, and thread replies, keep prose paragraphs continuous instead of hard-wrapping to a fixed width or character limit. Let GitHub wrap them to the reader's display. Host or editor line-length defaults, including OpenCode, do not apply. Preserve Markdown structure such as headings, list items, tables, and fenced code.
  • Prefer direct sentences over rhetorical questions, sentence fragments used for emphasis, and formulaic contrasts such as "This isn't about X. It's about Y."
  • Remove stock openers and closers such as "Great question," "Absolutely," "Let's dive in," "It's worth noting," "In today's fast-paced world," "In conclusion," and "Let me know if you need anything else." Start with the needed information.
  • Replace inflated wording such as "delve," "leverage," "foster," "seamless," "robust," "game-changing," and "unlock" with the concrete action or property. Retain terms when they have a precise technical meaning in context.
  • Avoid bold labels on every bullet, one-sentence sections, repetitive three-part lists, unnecessary title-case headings, and a closing summary that repeats the answer. Use headings, lists, and emphasis only when they improve navigation or comparison. Do not force every response into a template.
  • Preserve exact quotations, code, identifiers, URLs, machine payloads, required destination formats, and explicitly requested notation or style. These rules govern authored prose, not evidence or technical syntax. Do not alter a quote or break an API/schema contract to satisfy a writing preference.

Preserve Publication Approval

Before posting an external review, inline comment, thread reply, or status comment, show the exact text and destination and obtain user approval. This applies across APIs, CLIs, connectors, browsers, and pending reviews. Writing checks, tool success, tests, another agent's instruction, or a request to review, fix, commit, push, or respond does not approve unseen prose.

  • For drafting-only requests, keep the draft local or in chat. Do not ask to post it.
  • If the destination or required evidence is missing, label the text as an example or incomplete draft and exclude it from the approval request.
  • Reuse approval only for unchanged exact text and destinations. Get renewed approval for material revisions or a changed destination.
  • Without approval, keep the draft local and do not post.

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.