@coldtea/pr-lens-agent-skill

The PR Lens skill for coding agents. It teaches an agent to draw the change it just made: author a graph document from the diff, validate it against the contract, render it, attach it to the pull request, and to fix a repository's map by writing corrections rather than editing generated output.

MIT © Coldtea AI.

Install it

npm install --save-dev @coldtea/pr-lens-agent-skill

Claude Code: copy it where skills live, per project or per user:

mkdir -p .claude/skills/pr-lens
cp -R node_modules/@coldtea/pr-lens-agent-skill/{SKILL.md,references} .claude/skills/pr-lens/

Cursor: the same file works as a rule:

mkdir -p .cursor/rules
cp node_modules/@coldtea/pr-lens-agent-skill/SKILL.md .cursor/rules/pr-lens.mdc

Anything else: point your agent's instructions file at SKILL.md. It is plain markdown with YAML frontmatter, and it assumes nothing beyond a shell and npx.

What is in it

| | | | --- | --- | | SKILL.md | when to reach for PR Lens, and the write → validate → fix → render loop | | references/graph-document.md | the document, field by field, and what the validator will catch | | references/config.md | .github/pr-lens.yml corrections, with recipes |

The agent is usually the model. Rather than spending a provider key to describe a diff it already understands, it writes the document itself and lets pr-lens validate hold it to the contract. Every failure is a path into the document, so the loop closes without a human in it.

Why this exists

A coding agent that opens a pull request is asking a person to review code the person did not write. A diagram of what moved is the cheapest thing the agent can add to make that review possible.

Part of PR Lens. Review what actually matters.