| name | avoid-ai-writing |
| description | Audit and rewrite BoringStack docs and repo prose to remove AI writing patterns ("AI-isms"). Use for Starlight MDX under .github/src/content/docs, profile/README, root README, or when asked to remove AI-isms, clean up AI writing, or make copy sound less machine-generated. |
Avoid AI writing (audit and rewrite)
You are editing content to remove AI writing patterns ("AI-isms") that make text sound machine-generated.
The user will provide a piece of writing (or a path). Your job is to:
- Audit it: identify every AI-ism present, citing the specific text.
- Rewrite it: return a clean version with all AI-isms removed.
- Show a diff summary: briefly list what you changed and why.
BoringStack docs site (this monorepo)
When the scope is this documentation site or its siblings:
| Kind | Typical paths |
|---|
| Starlight pages | .github/src/content/docs/**/*.mdx |
| Site config copy | .github/astro.config.mjs (e.g. description, sidebar labels) |
| Org profile | .github/profile/README.md |
| Monorepo index | README.md at repo root |
| Deploy notes | .github/DEPLOY.md |
MDX frontmatter: title and description are YAML. If the rewritten description contains a colon followed by space (Key: value inside the string), wrap the whole value in double quotes so npm run build does not fail with a YAML parse error.
Do not break: Starlight component imports (Aside, Steps, Card, FileTree, etc.), code fences, tables, or mermaid blocks. Only change prose and safe frontmatter strings.
Verify when editing many pages: from .github/, run npm run build.
What to remove or fix
Formatting
- Unicode em dash (U+2014) and en dash abuse: Replace with commas, periods, parentheses, or two sentences. Target: zero em dashes in shipped docs. (CLI flags like
npm run build -- --help are fine.)
- Bold overuse: Strip bold from most phrases. At most one bolded phrase per major section, or none. Lead with the important word in plain text instead.
- Emoji in headers: Remove. No
## 🚀 What this means. Exception: social-only copy, one or two emoji at end of line, never mid-sentence.
- Excessive bullet lists: Turn bullet walls into prose where they are not true lists (steps, parameters, comparisons stay as lists).
Sentence structure
- "It's not X, it's Y" / "This isn't about X, it's about Y": Rewrite as a direct positive statement. At most one such contrast per piece, only if it earns its keep.
- Hollow intensifiers: Cut
genuine, real (as in "a real improvement"), truly, quite frankly, to be honest, let's be clear, it's worth noting that. State the fact.
- Hedging: Cut
perhaps, could potentially, it's important to note that, to be clear. Say the point.
- Missing bridge sentences: Paragraphs should connect; add a short bridge if blocks are interchangeable.
- Rule of three overload: Vary counts (two, four, or one full sentence). Cap triads of adjectives.
Structural issues
- Uniform paragraph length: Mix short punchy graphs with longer explanatory ones on purpose.
- Formulaic openings: Replace "In today's rapidly evolving…" style with the actual news or task.
- Suspiciously clean grammar: Keep deliberate fragments, sentences starting with "And" / "But," or light comma splices if they match a human voice.
For vocabulary tables and extra categories, see PATTERNS.md.
Output format
Return your response in four sections:
1. Issues found
Bulleted list of every AI-ism, with quoted offending text.
2. Rewritten version
Full rewritten content. Preserve structure, intent, and technical details. Change only what the guidelines require.
3. What changed
Short summary of meaningful edits (not a word-by-word ledger).
4. Second-pass audit
Re-read section 2. Note any surviving tells; fix and state what changed. If clean, say so.
Tone calibration
Aim for copy that sounds written by a careful engineer: direct, specific, confident without preening.
- Vary sentence length: mix short and long; fragments are fine.
- Be concrete: names, numbers, paths, commands, version hints.
- Have a voice: first person or clear editorial "we" where it fits BoringStack docs.
- Cut false neutrality: if the doc takes a stance (e.g. sibling repos vs monorepo), say it plainly.
- Earn emphasis: show why something matters; do not label it "interesting."
If the draft is already strong, say so and touch only what still violates the list above.
When to rewrite from scratch vs patch
If you see 5+ flagged vocabulary hits across categories, 3+ pattern categories, and flat rhythm, patching phrases will not work. Say so: one-sentence core message, then rebuild.
Developer audience (BoringStack docs)
All user-facing copy on the docs site assumes the reader ships code: repos, env vars, CI, Docker, HTTP, git.
- Lead with mechanism: file path, command, port, env var, or lint rule before any "why."
- Ban pitch patterns:
try it, company app, product spine, just works, payoff, noble, deadline-proof without citing a rule or command.
- Ban founder/PM framing on the homepage and generic intros:
first customers, token budget (keep agent-context language only on pages about AI workflows, e.g. architecture/agent-docs.mdx).
- Eyebrows are section labels, not slogans: prefer
Auth, Compose overlays, Stripe webhooks over Payment Spine, Ship shape, API product spine.
- CTAs are actions:
Quickstart, View on GitHub, Run locally, not Start the stack or Start with the bet.
- Fit-check tables (Why BoringStack): state technical constraints (runtime ownership, contract shape), not "good fit if you're building a SaaS."
- Do not count repos: say UI, API, infra, or "the templates". Never "three repos", "four templates", or similar counts that break when the org grows.
