Compound Engineering for Pi plugin
A loop of planning, work, review and recorded learnings for Pi, plus an extension with subagents and MCP access.
- 52 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The package starts subagents and child Pi processes with access to the project.
- Through `mcporter_call` the agent calls tools of connected MCP servers.
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/compound-engineering-for-piDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The package starts subagents and child Pi processes with access to the project.
- Through `mcporter_call` the agent calls tools of connected MCP servers.
This is third-party code. Review the repository files before installing.
What it does
The `compound-engineering-pi` package installs the Every Compound Engineering set into Pi: 84 skills, including the `ce:plan`, `ce:work`, `ce:review` and `ce:compound` workflows, reviewer personas and browser and video scenarios. Nine legacy prompts such as `/workflows-plan` and `/workflows-review` are kept for compatibility. The compatibility extension adds four tools: `ask_user_question`, `subagent` (`single`, `parallel` and `chain` modes), `mcporter_list` and `mcporter_call` for MCP access through MCPorter. A subagent can be given its own model, including the `provider/modelId` form.
Who it is for. Pi agent users who want the Compound Engineering plan, work and review loop without porting it by hand.
Good fit when
- A feature runs through the loop: `/skill:ce:plan` writes the plan, `ce:work` executes, `ce:review` does a multi-agent review.
- After a task you want to record learnings for later sessions, and `ce:compound` documents them.
- Work splits into independent chunks, and the `subagent` tool runs them in parallel or as a chain.
- Old habits stay: `pi -p "/workflows-plan Build a todo app"` still works.
- The agent has to call an MCP server, and in Pi `mcporter_call` does that once MCPorter is installed.
Not a fit when
- You work in Claude Code, where the original Compound Engineering plugin is installed.
- You need one small skill rather than a set of 84.
- You need richer subagents: with `pi-subagents` installed this package defers to it.
Example request
Make a plan for a todo app with `/skill:ce:plan`, then run `ce:review` on the finished code.Limitations
The package targets Pi and installation in other agents is not described. Content is synced from the main Compound Engineering plugin, so new skills arrive with a delay. A subagent model must be available and authenticated in your Pi installation, otherwise the default model of the child Pi process is used. MCP needs MCPorter installed separately with `npm i -g mcporter`. The last commit was in October 2026.
How to disable. Remove the package from Pi with `pi remove npm:compound-engineering-pi`, then run `/reload`.
FAQ
How do I install it in Pi?
With `pi install npm:compound-engineering-pi -l`, then `/reload` in the session.
Which workflows are included?
The Plan, Work, Review, Compound loop: `ce:plan`, `ce:work`, `ce:review` and `ce:compound` skills. Run them like `pi -p "/skill:ce:plan Build a todo app"`.
Do the old `/workflows-*` commands still work?
Yes, nine legacy prompts are kept for compatibility: `/workflows-plan`, `/workflows-work`, `/workflows-review`, `/workflows-compound` and others.
How does the subagent tool work?
It takes `single`, `parallel` and `chain` modes. A parallel run returns compact output by default; `includeOutputs: true` returns full output.
Where does the skill set come from?
It is the Compound Engineering plugin by Every and Kieran Klaassen. The repository covers Pi packaging, docs and the compatibility layer.
Discussion
No questions or reviews yet.
Related
PluginsAn extension for the Pi agent (pi.dev): subagents and orchestration in the Claude Code style, a widget and JavaScript scripts. Not for Claude Code
Personal Pi agent setup with four commands for plan, todos, execution and review, passing state through Markdown session files.
A minimal terminal coding agent extended with skills, prompt templates and TypeScript packages
An extension for the Pi agent: one programmable fabric_exec tool calls tools, MCP, agents and workflows from code
A set of extensions for the Pi agent: repository analysis, secrets leased with a TTL and search over agent sessions.
shitty-extensions adds fifteen extensions to the Pi agent: second opinions, plan mode, cost tracking and test loops.
README in short
The repository packages Compound Engineering skills for Pi and adds a compatibility extension with four tools. The main content is synced from the upstream plugin.