agentleFS
QuantConnectQuantConnect/Documentation/.claude/skills/docs-style/SKILL.md

docs-style

Skill268 starsChanged 48 days agoApache-2.0View on GitHub

Write, edit, or review QuantConnect documentation prose so it follows the repo Style Guide, covering scope, redundancy, contradictions, voice, tense, sentence length, word choice, headings, and HTML classes. Use when asked to "review the style of <page>", "check this page against the style guide", "copy-edit", "tighten the wording", "make it concise", "cut what we don't need", "standardize the voice", or when writing or rewriting any .html/.php page fragment under the numbered doc folders. Skip when the task is only code examples, metadata.json, generators, or page placement and renumbering.

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

31 rules. 0 are common, 25 are distinctive, and 6 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 25

    Distinctive

    This author's own choices. Worth a look.

  • 6

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Scope: Keep only what the reader needs to act on this page's topic.

    line 14
  2. Very specific:

    Contradictions: Check that numbers, dates, names, and claims agree across sections, the FAQ, metadata.json, related pages, and the sources.

    line 16
  3. Distinctive:

    Summaries: Don't add summary paragraphs that preview or recap sections.

    line 17
  4. Distinctive:

    Use active voice. Passive voice is fine only when the actor is unknown or unimportant, or when active voice would blame the reader.

    line 23
  5. Distinctive:

    Use present tense. Use *will* only for events that happen later from the reader's point of view, such as a dated migration.

    line 24
  6. Distinctive:

    Use verbs, not nominalizations: *describes*, not *provides a description of*.

    line 25
  7. Distinctive:

    Use *can* for ability, *may* for permission, *might* for possibility, and *must* for necessity.

    line 27
  8. Distinctive:

    Write plainly for a global audience: no idioms, metaphors, humor, or culture-specific references.

    line 28
  9. Distinctive:

    Use subject-verb-object order.

    line 32
  10. Distinctive:

    Keep sentences to 25 words or fewer.

    line 33
  11. Distinctive:

    Use a bullet list for three or more parallel items.

    line 35
  12. Distinctive:

    Don't start a sentence with *It is*, *There is/are*, or a bare *This*.

    line 36
  13. Distinctive:

    Use the shortest correct term, after you apply the Content rules.

    line 37
  14. Distinctive:

    Use one term per concept across the page and its siblings.

    line 38
  15. Distinctive:

    Every sentence carries context.

    line 44
  16. Distinctive:

    Don't use a colon to set up a punchline or a reveal.

    line 46
  17. Distinctive:

    Leave out editorial commentary and advice about attitude, such as *which is the point of the change*, *the upside is*, and *treat this as*.

    line 47
  18. Distinctive:

    Only claim a cause, a frequency, or a *most common* outcome when the data shows it.

    line 48
  19. Distinctive:

    When you're asked to embed or link a source, such as a report, embed or link it.

    line 50
  20. Distinctive:

    Use the serial comma. End every sentence and step with a period.

    line 54
  21. Distinctive:

    Use periods, not semicolons.

    line 55
  22. Distinctive:

    Don't use a slash for a choice, except in established terms like *read/write*.

    line 56
  23. Distinctive:

    Follow every heading with body text, and don't start that text with a pronoun that refers to the heading.

    line 62
  24. Very specific:

    Put a member that has a Python name in both languages: <code class="csharp">PERatio</code><code class="python">pe_ratio</code>.

    line 66
  25. Very specific:

    Put code blocks in <div class="section-example-container"> with <pre class="python"> and <pre class="csharp">.

    line 67
  26. Very specific:

    Use table qc-table for tables and introduce each table with a sentence that ends in a colon.

    line 68
  27. Distinctive:

    Keep terms that are standard in LEAN docs even when the guide lists them as jargon.

    line 70
  28. Distinctive:

    Run the linter on the page folder or files.

    line 74
  29. Very specific:

    Read the whole page, including the FAQ, metadata.json, and any embedded source, before you judge single sentences.

    line 80
  30. Very specific:

    Read every file in full for what the linter can't detect: voice, humanized software, nominalizations, *should*, terminology drift between …

    line 81
  31. Distinctive:

    When you apply fixes, keep the facts, numbers, and links exactly as they are.

    line 83

“Common” is measured: kept by at least 6 in 10 of the 3 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

Documentation Style and Voice

The source of truth is Style Guide/. This skill condenses it and adds the house conventions the guide does not cover. Read the guide file named in a rule when you need its full table.

Content

Being concise means leaving out what the reader doesn't need, not using fewer words. Before you edit any wording, decide which sentences belong on the page.

  • Scope: Keep only what the reader needs to act on this page's topic. Link to background, history, internals, and related topics instead of explaining them. Cut statistics that don't change what the reader does.
  • Redundancy: State each fact once. Remove any sentence that repeats another section, the page introduction, an FAQ answer, a table on the page, or an embedded source such as an iframe or a report. Link to another page that already covers the fact instead of restating it.
  • Contradictions: Check that numbers, dates, names, and claims agree across sections, the FAQ, metadata.json, related pages, and the sources. When two statements conflict, ask the owner which one is correct. Don't pick one yourself.
  • Summaries: Don't add summary paragraphs that preview or recap sections. The headings already give the page structure.

Voice

  • Address the reader as you, and use the imperative for instructions. Say we recommend, not it's recommended or QuantConnect recommends.
  • Phrase guidance positively. Tell the reader what to do instead of what to avoid, including in warnings and common-mistake lists.
  • Use active voice. Passive voice is fine only when the actor is unknown or unimportant, or when active voice would blame the reader.
  • Use present tense. Use will only for events that happen later from the reader's point of view, such as a dated migration.
  • Use verbs, not nominalizations: describes, not provides a description of. Use single-word verbs: determine, not figure out.
  • Software can detect, store, require, or process. It does not know, want, remember, or see.
  • Use can for ability, may for permission, might for possibility, and must for necessity. Avoid should unless you explain why.
  • Write plainly for a global audience: no idioms, metaphors, humor, or culture-specific references. See style-guide-terminology-global-audience.md.

Sentences and paragraphs

  • Use subject-verb-object order. Put the condition or location before the action: If X, do Y. In the panel, click Z.
  • Keep sentences to 25 words or fewer. Split a longer sentence at its natural clause break.
  • Cover one idea per paragraph, in two to five sentences.
  • Use a bullet list for three or more parallel items. Use a numbered list for steps, with one action per step.
  • Don't start a sentence with It is, There is/are, or a bare This. Name the real subject or follow this with a noun.
  • Use the shortest correct term, after you apply the Content rules. The concise terms table lists replacements: to, not in order to. Must, not need to or have to. And, not as well as.
  • Use one term per concept across the page and its siblings. See style-guide-consistent-terminology.md.

Plain prose, not AI prose

Reviewers reject text that reads as AI-generated. Write connected explanatory sentences that give the reader the context they need.

  • Every sentence carries context. Replace clipped openers such as Backtest results change. or The differences are not cosmetic. with a sentence that says what changes and why.
  • State what is true. Avoid contrast framing such as X, not Y, not A, it is B, and a reason to X, not a bug to Y.
  • Don't use a colon to set up a punchline or a reveal.
  • Leave out editorial commentary and advice about attitude, such as which is the point of the change, the upside is, and treat this as.
  • Only claim a cause, a frequency, or a most common outcome when the data shows it. When you don't know, describe the mechanism and stop.
  • Take dates, timelines, and availability from the owner of the change. Don't infer them from commit dates or data.
  • When you're asked to embed or link a source, such as a report, embed or link it. Don't write a summary in its place.

Punctuation

  • Use the serial comma. End every sentence and step with a period.
  • Use periods, not semicolons. Use a colon only to introduce a list, table, code block, or example.
  • Don't use a slash for a choice, except in established terms like read/write. Don't use exclamation marks. Use question marks only in FAQ questions.

Headings

  • House convention: page file names and <h4> headings use Title Case, such as 04 Reporting Dates.html and <h4>Update Frequency</h4>. This overrides the guide's sentence case.
  • Match the heading form to the content. Use a noun phrase for concepts and references (Coverage, Retired Data Points). Use an imperative verb for tasks (Deploy Live Algorithms). End an FAQ heading with a question mark.
  • Follow every heading with body text, and don't start that text with a pronoun that refers to the heading. Don't create a single sub-heading at any level.

HTML conventions

  • Put a member that has a Python name in both languages: <code class="csharp">PERatio</code><code class="python">pe_ratio</code>. The same applies to true/True and null/None. Use a single <code> only for names that are identical in both languages, such as class and exception names.
  • Put code blocks in <div class="section-example-container"> with <pre class="python"> and <pre class="csharp">. A plain <pre> is only for output that is the same in both languages.
  • Use table qc-table for tables and introduce each table with a sentence that ends in a colon.
  • Mark UI and file elements with the classes in style-guide-html-classes.md: button-name, field-name, menu-name, tab-name, public-file-name, new-term, and others.
  • Keep terms that are standard in LEAN docs even when the guide lists them as jargon. For example, keep throws an exception.

Review workflow

  1. Run the linter on the page folder or files. It reads the Avoid lists from the Style Guide tables:

    python .claude/skills/docs-style/lint.py "<page folder>"
  2. Read the whole page, including the FAQ, metadata.json, and any embedded source, before you judge single sentences. Check the page against the Content rules. List every out-of-scope, redundant, or contradicting passage.
  3. Read every file in full for what the linter can't detect: voice, humanized software, nominalizations, should, terminology drift between sections, dual-language <code>, and heading form.
  4. Report the findings. Put contradictions first, then redundant and out-of-scope passages, then terminology and markup, then wording. For each finding, quote the text, name the rule, and give the fix, which is often a cut.
  5. When you apply fixes, keep the facts, numbers, and links exactly as they are. Edit only the sentences you flagged, and keep the file's existing line layout.

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.