Write and review Markdown (.md) files with consistent structure, correct links, and repo-friendly conventions. Use when editing README, ADRs, AGENTS.md, llms.txt, SKILL.md, copilot-instructions.md, or any documentation page.
Purpose: help agents reliably author and review Markdown with correct structure, links, and minimal ambiguity. This skill is intended to be used by any agent that creates/edits/reviews
.mdcontent.
Load this skill when you:
*.md file.github/decisions/AGENTS.md, llms.txt, or copilot-instructions.mdSKILL.md packages or .agent.md prompts (Markdown structure still applies)Agents writing or editing Markdown in this repo must satisfy all of the following:
# ... heading.# → ## → ### (no skipped levels)./ as a path separator (never \).yaml, json, bash, text).--- lines.- for unordered lists.# <Title>
## Purpose
<What this document is for, in 1–3 sentences>
## Audience
<Who should read this>
## Workflow / Process
1. <Step>
2. <Step>
## References
- <links>
# ADR-XXXX — <Title>
## Status
Proposed | Accepted | Superseded by ADR-YYYY
## Context
<Why we needed to decide>
## Decision
<What we decided, one sentence>
## Consequences
<What changes; what becomes easier/harder>
# title at top; headings are unique and ordered./.Use these when you want a repeatable “doc QA” pass. If the repo does not have these tools configured, treat this section as guidance (do a manual check instead).
Suggested tooling (choose one set; do not require all):
markdownlint-cli2 (Markdown lint)cspell (spell check)lychee (link check)Minimum manual equivalent if you cannot run tools:
#, no skipped levels, no duplicates.framework/spec/appendices/01-appendix-a1-ai-and-llm-standards.md (A1.1)