CCPM – The Project Manager Agent
Spec-driven development for AI agents: turn ideas into PRDs, GitHub issues, and parallel-executed code with full traceability.
- Source repo
- automazeio/ccpm
- Stars
- ★ 8.4k
- Last updated
- 7mo ago
- License
- MIT
- Primary language
- Shell
- FA score
- 47/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex · Claude Code
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- An indie developer who wants to turn a rough idea into a structured spec before coding, avoiding context loss across sessions.
- Not a fit if
- Teams not on GitHub: it requires GitHub Issues and the gh CLI
- Small projects or one-off scripts: the PRD/epic/task process is heavyweight
- Source review
- 47/100 · Major gaps 2 safety controls not found
What does this agent do, and when should you use it?
CCPM is an Agent Skill that brings a structured project management brain to AI coding agents. It uses GitHub Issues as the single source of truth and Git worktrees for parallel execution. The workflow spans brainstorming, PRD creation, technical epic generation, task decomposition, GitHub sync, and parallel agent execution. All status and tracking operations are deterministic bash scripts, ensuring fast, consistent output without LLM token costs. CCPM is harness-agnostic, following the agentskills.io open standard and working with Claude Code, Codex, OpenCode, Factory, Amp, Cursor, and more. It eliminates context loss, provides full traceability, and enforces a 'no vibe coding' discipline where every line of code traces back to a specification.
CCPM begins with guided brainstorming, producing PRD files in .claude/prds/. It then parses PRDs into technical epics stored under .claude/epics/<feature>/epic.md. Epics are decomposed into task files with metadata like depends_on, parallel, and conflicts_with. The sync step uses GitHub CLI to create an epic issue, sub-issues, and a dedicated worktree. Execution analyzes an issue for independent work streams, launches parallel agents scoped to separate files, and tracks progress via Git commits. Tracking commands like status.sh, standup.sh, and search.sh scan project files and report instantly.
- An indie developer who wants to turn a rough idea into a structured spec before coding, avoiding context loss across sessions.
- A team that needs to parallelize feature work across multiple agents without conflicts, using git worktrees for isolation.
- A tech lead who wants to automatically decompose a PRD into a technical epic and then into actionable GitHub issues.
- Organizations that require full traceability from requirements to code, with every decision documented in issues and commits.
- Teams that want to use GitHub Issues as the central collaboration hub, enabling seamless handoffs between human and AI agents.
How do you install or deploy this agent?
Clone the repository: git clone https://github.com/automazeio/ccpm.git. For Factory/Droid, symlink the skill: ln -s /path/to/ccpm/skill/ccpm ~/.factory/skills/ccpm. For Claude Code, create a .claude/skills/ directory in your project root and symlink: ln -s /path/to/ccpm/skill/ccpm .claude/skills/ccpm. Ensure Git and GitHub CLI are installed and authenticated (gh auth login). Install the gh-sub-issue extension: gh extension install yahsan2/gh-sub-issue.
How do you use this agent?
After installation, CCPM automatically activates when you express PM intent. Speak naturally: "I want to build a notification system". CCPM will ask clarifying questions and create a PRD. Then say "parse the notification-system PRD" to create an epic, "break down the notification-system epic into tasks" to decompose, "sync the notification-system epic to GitHub" to sync issues and worktree, and "start working on issue 42" to launch parallel agents. Use "standup", "what's next", and "what's blocked" to get instant status.
What are this agent's strengths and limitations?
- Enables true parallel execution via git worktrees, potentially up to 3× faster feature delivery.
- Full traceability from PRD → Epic → Task → Issue → Code → Commit.
- Deterministic bash scripts for status and standup, providing instant output with zero LLM cost.
- Harness-agnostic, following the agentskills.io standard; works across multiple tools.
- GitHub Issues as source of truth, supporting team collaboration and handoffs.
- Requires a GitHub repository and GitHub CLI setup, coupling to GitHub infrastructure.
- Steep learning curve to understand the PRD/epic/task/sync workflow.
- Parallel execution can introduce Git conflicts; mitigation requires discipline.
- Performance claims (89% less context switching, 75% bug reduction) are self-reported and lack independent validation.
- May be overkill for small projects; the file structure and process may feel heavyweight.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| CCPM – The Project Manager Agent This agent | 47 · Major gaps | Agent plugin / skillFree + model costs | ★ 8.4k | 7mo ago | Shell | Codex · Claude Code |
| Fusion Software Factory | 71 · Some gaps | CLIFree + model costs | ★ 1.3k | 1d ago | TypeScript | Codex · OpenAI API · Claude API |
| Keep the Why | 75 · Good | Agent plugin / skillFree | ★ 231 | 1d ago | Python | Codex · Claude Code |
| dmux | 69 · Some gaps | CLIFree + model costs | ★ 1.8k | 2mo ago | HTML | Codex · Claude Code · OpenAI API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence: README describes GitHub integration (gh CLI, gh-sub-issue extension) and local file operations, but does not explicitly state least-privilege principles; user confirmation exists (e.g., before sync) but not systematic; data flow transparency partially (file paths, GitHub sync) but lacks sensitive data handling details; dependency security not mentioned; external effects (creating issues, worktrees) described but no rollback mechanism; source attribution (author, company) mentioned but publisher unverified. Deductions: least privilege, user confirmation, data flow transparency, external effects, rollback, source attribution only partial; sensitive data handling and dependency security absent.
Evidence: README internally consistent, workflow clear; dependencies (git, gh, gh-sub-issue) listed but no versions or installation verification; failure messages not mentioned. Deductions: self-consistency good, but dependency availability and failure messages insufficient.
Evidence: target audience (developers, teams) and scenarios (PRD, epic, parallel execution) clear; capability boundaries (local vs GitHub) table; trigger words (natural language) list; environment fit (multiple harnesses) mentioned. Deductions: capability boundaries and trigger precision partial, environment fit lacks specific configuration details.
Evidence: information architecture clear (TOC, structure diagram); install notes detailed (clone, symlink); naming stability explained (task file naming); examples and FAQ have example flow but no FAQ; known limitations not explicit; license MIT; versioning/changelog not mentioned; maintenance responsibility (author, company) mentioned. Deductions: known limitations and versioning/changelog missing, maintenance responsibility partial.
Evidence: output usability (standup, status reports) described; marginal value (parallel execution, context preservation) explained; cost-benefit (reduced context switching, bug rates) claimed. Deductions: output usability, marginal value, cost-benefit based on claims, lacking independent verification.
Evidence: claims (e.g., 89% reduction) lack traceable sources; benchmark results (100% vs 27.7%) lack test details; facts and inferences not clearly separated. Deductions: claims lack traceability, cross-source corroboration insufficient, facts and inferences mixed.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Publisher identity unverified; treat source with caution.
- Dependency security not mentioned; check security of gh-sub-issue extension before use.
- Sensitive data handling not specified; review before handling credentials or private data.
- Rollback mechanism unclear; backup before sync or merge operations.
- Performance claims (e.g., 89% reduction) lack verifiable evidence; do not rely on them for decisions.
FAQ
Is CCPM only for Claude Code?
What are the prerequisites?
gh-sub-issue extension is required for parent-child issue relationships, though CCPM falls back to task lists if not installed.How does CCPM handle parallel agent conflicts?
conflicts_with to prevent overlapping changes.