Registry / agentpack
Pull Request Quality Pack
Cross-platform PR review, security review, formatting, and summary workflow.
#code-review#pull-requests#security
Unsignedcritical
v0.1.0 · agentpackInstall
Validate
npx agentpack validate examples/pr-quality
Install (writes to your project, with diff + lockfile)
npx agentpack install examples/pr-quality --target claude-code --profile safe
Export only (no install, writes to dist/)
npx agentpack pack export examples/pr-quality --target claude-code --profile safe --out dist/claude-code
Verify (drift detection)
npx agentpack verify agentpack.pr-quality
From your project root. install prints a diff and prompts before writing; pass --dry-run to preview, --yes to skip the prompt, or --force to overwrite files without an AgentPack marker. The lockfile (AGENTPACK.lock) is committed; everything else under .agentpack/ is per-machine.
Profiles
- safe — Instruction, rule, skill, and command only. No shell hooks, MCP, secrets, or external APIs.
- standard — Adds the security reviewer subagent. Still no hooks, MCP, or secrets.
- full — Adds formatting hook and optional GitHub MCP connector.
- enterprise — Full install with signature, lockfile, and admin approval required.
Compatibility
| Target | Status | Notes |
|---|---|---|
| Claude Code | Supported | Generates CLAUDE.md, .claude/skills, .claude/agents, and .claude/settings.json. |
| Codex | Supported | Generates AGENTS.md, .codex/config.toml, .codex/hooks.json, and skills. |
| Cursor | Supported | Generates .cursor/rules, .cursor/mcp.json, and AGENTS.md. Hook output is conservative. |
| ChatGPT Apps | Experimental | Generates MCP app skeleton and project-instructions.md only. |
| Generic / AGENTS.md | Supported | Generates AGENTS.md, skills/, README-agent.md, and agentpack.json. |
Supported = the adapter emits complete output for that target; how the platform consumes it varies (e.g. Codex reads repo-root AGENTS.md, not project .codex/ files) — see adapter docs.
Permission summary
Profile: safe
DefaultLow risk
- filesystem.read — Read files in the project(from skill:code-review, command:pr-summary)
- git.operations — Run git read operations (status/diff/log)(from command:pr-summary)
Profile: standard
Low risk
- filesystem.read — Read files in the project(from skill:code-review, command:pr-summary, subagent:security-reviewer)
- git.operations — Run git read operations (status/diff/log)(from command:pr-summary)
Profile: full
High risk
- filesystem.write — Write or modify files in the project(from hook:post-edit-format)
- mcp.server — Run or configure an MCP server(from mcp_server:github)
- repo.modification — Modify repository contents(from mcp_server:github)
- secrets.env — Read environment variables / secrets(from mcp_server:github)
- shell.execution — Run shell commands on your machine(from hook:post-edit-format)
Medium risk
- external_api.access — Call an external HTTP API(from mcp_server:github)
- network.access — Make outbound network requests(from mcp_server:github)
Low risk
- filesystem.read — Read files in the project(from skill:code-review, command:pr-summary, subagent:security-reviewer)
- git.operations — Run git read operations (status/diff/log)(from command:pr-summary)
Secrets required
- GITHUB_TOKEN — Required only for the optional GitHub MCP server.
Network domains
- api.github.com
Declared shell commands
- $ npm run format
- $ npm run lint
Profile: enterprise
High risk
- filesystem.write — Write or modify files in the project(from hook:post-edit-format)
- mcp.server — Run or configure an MCP server(from mcp_server:github)
- repo.modification — Modify repository contents(from mcp_server:github)
- secrets.env — Read environment variables / secrets(from mcp_server:github)
- shell.execution — Run shell commands on your machine(from hook:post-edit-format)
Medium risk
- external_api.access — Call an external HTTP API(from mcp_server:github)
- network.access — Make outbound network requests(from mcp_server:github)
Low risk
- filesystem.read — Read files in the project(from skill:code-review, command:pr-summary, subagent:security-reviewer)
- git.operations — Run git read operations (status/diff/log)(from command:pr-summary)
Secrets required
- GITHUB_TOKEN — Required only for the optional GitHub MCP server.
Network domains
- api.github.com
Declared shell commands
- $ npm run format
- $ npm run lint
Atoms
| Atom | Type | Risk | Description |
|---|---|---|---|
| instruction:pr-review-standards | Instruction | low | Persistent standards for reviewing pull requests. |
| rule:security-review-required | Rule | low | Flags security-sensitive code changes for explicit review. |
| skill:code-review | Skill | low | Reusable pull request review procedure with quality, maintainability, bug, and test coverage checks. |
| command:pr-summary | Command | low | Generates a clear PR summary, risk notes, and testing notes. |
| subagent:security-reviewer | Subagent | medium | Specialized security review agent for auth, secrets, input validation, permissions, and data exposure. |
| hook:post-edit-format | Hook | high | Runs project formatter after agent file edits. |
| mcp_server:github | MCP server | high | Optional GitHub MCP connector for PR metadata, comments, issue links, and repository context. |
Manifest preview
AGENTPACK.yaml
agentpack: "1.0"
metadata:
id: "agentpack.pr-quality"
name: "Pull Request Quality Pack"
slug: "pr-quality"
description: "Cross-platform pull request review workflow with code review, security review, formatting, and PR summary generation."
version: "0.1.0"
license: "MIT"
publisher: "agentpack"
authors:
- name: "AgentPack"
email: "hello@agentpack.dev"
tags:
…Validation
✓ Manifest validates clean.