CCPM
Project-management skill system for AI agents — GitHub Issues plus Git worktrees for parallel execution.
About
CCPM (automazeio/ccpm) is a single Agent Skill that turns an AI coding agent into a project manager: a five-phase pipeline goes PRD → epic → numbered task files → GitHub Issues → parallel agents executing in a shared epic worktree. Project state lives in frontmatter-schema'd markdown under .claude/prds/ and .claude/epics/, so context survives across sessions; GitHub Issues become the shared status surface after sync, and 14 shipped bash scripts handle deterministic ops — status, standup, next, blocked, validate — with zero LLM token cost.
It follows the agentskills.io open standard, so it works harness-agnostic across Claude Code, Factory, Amp, OpenCode, Codex, and Cursor, and it holds roughly 8,000 stars. Honest caveats: it requires git and an authenticated gh CLI plus a GitHub repo — no GitHub, no CCPM — and the artifact-heavy workflow (per-issue analysis files, per-stream progress trackers) can feel bureaucratic for small one-shot tasks.
What it can do
- PRD-to-epic-to-Issue delivery pipeline
- Parallel agent execution in shared worktrees
- File-based project state with full traceability
- 14 deterministic bash-script ops
- Harness-agnostic via agentskills.io standard
Setup
Clone the repo and symlink skill/ccpm into your client's skills directory (e.g. .claude/skills/ccpm); requires git and gh CLI authenticated.
Prerequisites
git and gh CLI (gh auth login); a GitHub repository; an Agent Skills-compatible harness.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 66; Trending score 80/100 from researcher assessment.
★ 8,000 GitHub stars
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.