[Style Guide] Add rules against LLM-written content patterns#31269
[Style Guide] Add rules against LLM-written content patterns#31269mvvmm wants to merge 3 commits into
Conversation
|
This pull request requires reviews from CODEOWNERS as it changes files that match the following patterns:
|
703ea3a to
54ffadc
Compare
There was a problem hiding this comment.
1 issue needs fixing before merge.
CRITICAL — and contain unrelated dependency bumps (astro, vitest, wrangler, typescript-eslint, etc.) that are not mentioned in the PR title or description. Remove these files from the PR to avoid scope creep and unintended side effects.
LOW — Posted 1 inline suggestion for grammar consistency in the agent reference.
Out of scope — The PR description lists many LLM anti-patterns, but the human-facing style guide only reflects the emoji-as-admonition rule. Consider whether rules like "no throat-clearing introductions" and "avoid condescending qualifiers" should also appear in the human-facing writing guidelines.
| - **Filler qualifiers.** "Simply", "just", "easily", "quickly" — remove them. They add no information and imply the reader is at fault if they struggle. | ||
| - **Throat-clearing intros.** "In this guide, you will learn..." and "By the end of this tutorial..." restate the title. Delete them. | ||
| - **Closing summaries.** Paragraphs that restate what was just written. Delete them. | ||
| - **"We" and "let's".** Use imperative ("Run the command") or second person ("You can configure..."). Not "Let's run" or "We will configure". |
There was a problem hiding this comment.
Fix grammar to match imperative style of surrounding bullets:
| - **"We" and "let's".** Use imperative ("Run the command") or second person ("You can configure..."). Not "Let's run" or "We will configure". | |
| - **Use em dashes sparingly.** An occasional em dash for a sharp aside is fine; using them in every other sentence is a sign the sentence should be restructured. Prefer a period or a comma. |
|
Correction to my review above: The filenames got mangled by shell escaping. The files that need to be removed from this PR are |
|
I've completed the review for PR #31269. Labels applied: Summary of changes
Issues flagged
|
|
Preview URL: https://54ffadcb.preview.developers.cloudflare.com Files with changes (up to 15) |
|
/bonk re-review |
| - No emojis in page titles, headings, or sidebar labels. | ||
| - **No emojis anywhere in docs.** No emoji in prose, titles, headings, sidebar labels, or tables. Do not use emoji as admonition substitutes — `💡 Tip:` and `⚠️ Warning:` in prose are not supported patterns. Use `:::tip` and `:::caution` instead. | ||
| - **Avoid LLM-written patterns.** These are the most common tells — flag and fix them: | ||
| - **Em dashes overused.** One per page is a sign of good writing. One per paragraph is not. Restructure or use a comma or period. |
There was a problem hiding this comment.
Use imperative mood to match the surrounding bullets:
| - **Em dashes overused.** One per page is a sign of good writing. One per paragraph is not. Restructure or use a comma or period. | |
| - **Use em dashes sparingly.** One per page is a sign of good writing. One per paragraph is not. Restructure or use a comma or period. |
|
Posted the re-review on PR #31269. What I found:
|
Summary
Adds explicit rules to the agent style guide reference and the human-facing admonitions page covering patterns that appear frequently in LLM-generated docs:
💡 Tip:,⚠️ Warning:) instead of:::note/:::caution/:::tipdirectivesDocumentation checklist