AGENT SKILL

documentation-style

Formatting and style rules for markdown documentation. Use when generating or editing any project markdown document. Do not use for source code, code comments, or commit messages. For work item content (titles, descriptions, comments on GitHub Issues or Azure DevOps), use the Content Style section of the `work-item-creation` skill.

使用场景文档处理
License 许可
MIT
最近更新
2026-11-02
Stars
★ 50
Forks
12

01 前言 Front Matter

name
documentation-style
description
Formatting and style rules for markdown documentation. Use when generating or editing any project markdown document. Do not use for source code, code comments, or commit messages. For work item content (titles, descriptions, comments on GitHub Issues or Azure DevOps), use the Content Style section of the `work-item-creation` skill.

03 SKILL.md 正文

Documentation Style Guide

This guide defines standards for all documentation generated in the project.

Formatting

  • Ensure there are no spelling errors. Avoid emojis and decorative Unicode characters (such as →, •, ★, ✓, ?, ✅).
  • Do not introduce the em-dash character (Unicode U+2014) in newly written or rewritten Markdown text. Rephrase the sentence instead.
  • Avoid semicolons in newly written or rewritten prose. Recast the sentence naturally using a conjunction, a clearer sentence structure, or separate sentences when appropriate. Preserve the original meaning, and avoid turning connected ideas into a series of short, abrupt sentences. Keep semicolons required by code, technical syntax, or verbatim quotations.
  • Preserve existing text outside the requested scope. Do not scan or rewrite documents solely to remove existing em dashes or semicolons.
  • Avoid hyphens or dashes as separators between concepts. Instead of "concept A - concept B", rewrite the sentence.
  • Prefer lists and tables over long paragraphs.
  • Keep documents concise: specs and ADRs should be 1-2 pages, not novels.
  • Never use #<number> in free text (e.g., "#1 priority", "#3 option"). Azure DevOps automatically converts #<number> into a work item link. Use "first", "third option", or rephrase the sentence.

Language

  • Avoid promotional or exaggerated language.
  • Do not use contrastive constructions:
    • "it's not just X, it's Y"
    • "it's not just about..."
    • "more than just..."
    • "goes beyond..."
  • Describe directly what something is, without negating or minimizing alternatives.
  • Use active voice and direct sentences.
  • Prefer plain, widely understood vocabulary so non-native English speakers can follow the document. Use "place" or "location" instead of "locus", "use" instead of "utilize", and "make easier" instead of "facilitate" when the meaning is unchanged. Keep precise domain terms, API names, code identifiers, and verbatim quotations. Explain necessary unfamiliar terms on first use instead of replacing them with a less accurate word.
  • Avoid generic framing: if the sentence works in any project without modification, it is not saying anything useful.
    • Avoid: rhetorical questions followed by an obvious answer ("The solution? Test everything that matters.")
    • Prefer: go straight to the statement with project-specific context.

Names and References

  1. Use descriptive headings, scenario names, and plain statements instead of inventing codes for requirements, stories, tasks, assumptions, risks, or review findings. Do not replace numeric codes with another opaque naming scheme.
  2. Add formal identifiers only when explicitly requested or required by an established project convention, compliance process, or artifact-specific template or instruction. Current spec templates and verification rules still require requirement and conformance IDs, so preserve those contracts until they are changed together.
  3. Reference behavior by its descriptive name and document or section link when available. Use a precise link when names could be ambiguous. Preserve native board IDs, existing ADR numbers, and technical identifiers such as protocol fields and error codes.
  4. Keep ordinary numbered procedures and meaningful numeric values. Avoid repeating the same requirement in multiple sections solely to assign it another identifier, while retaining distinct inputs, expected outcomes, and constraints.
  5. Apply these defaults to newly written or rewritten text only. Preserve existing identifiers, anchors, and inbound references outside the requested scope. Convert an existing document only within an approved scope that includes its affected references.

Diagrams

For diagram design, review, tool selection (Mermaid vs Draw.io), and platform rendering rules, use the diagram-design skill.

Document Structure

  • Start with the most important content (conclusion, decision, summary).
  • Use consistent hierarchical headings (H1 for title, H2 for main sections).
  • Include metadata when relevant (Status, Date, Version).

04 同仓库更多 Skills

microsoft
microsoft/devsquad-copilot

work-item-workflow

Work item workflow for implementation. Use when starting work on a GitHub issue or Azure DevOps work item to verify assignee, dependencies, priority, and capacity. Do not use for creating new work items (use work-item-creation), for suggesting the next task (use next-task), or when the work source is only tasks.md without an issue/work item.

Skill与智能体开发
★ 50
microsoft
microsoft/devsquad-copilot

work-item-creation

Rules and checklist for creating work items on GitHub Issues or Azure DevOps. Use when you need to create issues, user stories, tasks, epics, or features on the board. Do not use for platform configuration (use board-config), managing workflow of existing tasks (use work-item-workflow), or for comments on work items.

★ 50
microsoft
microsoft/devsquad-copilot

triage-workflow

Label-based state machine for triaging issues and work items. Manages state transitions, generates durable triage artifacts, and supports resumable sessions. Use when triaging GitHub issues or Azure DevOps work items. Do not use for creating new work items from scratch (use work-item-creation) or for managing implementation workflow of already-triaged items (use work-item-workflow).

Skill与智能体开发
★ 50
microsoft
microsoft/devsquad-copilot

test-discipline

Testing practices for behavior-focused, refactor-resistant tests. Use when implementing features, fixing bugs, refactoring, or adding or reviewing tests to ensure they verify intent through public interfaces and avoid tautological assertions. Also use when adding regression tests after bug fixes. Consulted by the implement agent during coding and by the review agent when validating test quality. Do not use for test strategy decisions (use complexity-analysis), for failure triage and root-cause analysis (use debugging-recovery, then after root cause is localized use test-discipline to add the reproducing regression test and fix), or for integration test infrastructure setup.

测试与质量
★ 50
microsoft
microsoft/devsquad-copilot

security-review

Security assessment workflow in two modes: architectural (design) and code (implementation). Use when a security trigger is detected during planning (architectural mode) or implementation/review (code mode). Covers STRIDE, OWASP, dependency scanning, Azure compliance, and GitHub security alerts. Do not use for general code quality (use devsquad.review), for threat modeling as a standalone activity, or for compliance audits.

视觉设计
★ 50
microsoft
microsoft/devsquad-copilot

reasoning

Record decisions and justifications during SDD agent execution, and pass structured context between agents via handoff envelope. Use when an agent needs to document decisions made (Reasoning Log) or perform a handoff to another agent. Do not use for architecture documentation (use ADRs), complexity analysis (use complexity-analysis), or debug logs.

★ 50