@codersbrew/pi-tools
A pi package bundling CodersBrew pi extensions and skills.
Package details
Install @codersbrew/pi-tools from npm and Pi will load the resources declared by the package manifest.
$ pi install npm:@codersbrew/pi-tools- Package
@codersbrew/pi-tools- Version
0.7.0- Published
- Apr 5, 2026
- Downloads
- 483/mo · 46/wk
- Author
- codersbrew
- License
- MIT
- Types
- extension, skill, prompt
- Size
- 147.7 KB
- Dependencies
- 0 dependencies · 5 peers
Pi manifest JSON
{
"extensions": [
"./extensions"
],
"skills": [
"./skills"
],
"prompts": [
"./extensions/subagent/prompts"
]
}Security note
Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.
README
@codersbrew/pi-tools
A publishable pi package that bundles CodersBrew pi extensions, skills, and prompt templates.
Included resources
Extensions
security
Protects common dangerous tool operations by:
- warning or blocking risky
bashcommands such asrm -rf,sudo, and destructive disk operations - blocking writes to sensitive paths like
.env,.git,node_modules, SSH keys, and common secrets files - prompting before lockfile edits such as
package-lock.json,yarn.lock, andpnpm-lock.yaml
session-breakdown
Adds an interactive TUI for analyzing pi session history from ~/.pi/agent/sessions, including:
- sessions, messages, tokens, and cost over the last 7 / 30 / 90 days
- model, cwd, day-of-week, and time-of-day breakdowns
- contribution-style heatmap visualizations
update-pi
Adds a /update-pi slash command that upgrades @mariozechner/pi-coding-agent using your detected package manager (or configured npmCommand) and reminds you to restart pi afterward.
subagent
Delegates work to specialized subagents that run in isolated pi subprocesses.
Primary bundled agents:
planner- investigates the codebase and produces tracked plansexecutor- materializes, executes, resumes, and updates tracked plansreviewer- performs an independent read-only review pass
Internal helper:
worker- delegated implementation helper used by the executor
Primary prompt templates:
/plan/execute-plan/implement/implement-and-review/review
Tracked markdown plans are written to plan/*.md with task IDs, status checkboxes, dependency metadata, and validation steps. Executors can fan out safe parallel worker batches, resume interrupted work, and update shared plan files as work completes.
Override behavior:
- packaged agents act as defaults
~/.pi/agent/agents/*.mdoverrides packaged agents of the same name.pi/agents/*.mdoverrides both packaged and user-level agents whenagentScope: "project"or"both"
Skill: github-workflow
An opinionated GitHub workflow skill for branch creation, commits, pushes, PR creation, review, and merge work.
It prefers gh for GitHub-aware actions and uses plain git where local source-control operations are the better tool.
Invoke it in pi with:
/skill:github-workflow
Skill: plan
A /plan-style planning skill that investigates the codebase and writes a tracked markdown plan under plan/, but does the work in the current conversation context instead of delegating to the bundled subagent workflow.
It is useful when you want the same planning output shape as /plan without spinning up a separate planner/executor process.
Invoke it in pi with:
/skill:plan add Redis caching to the session store
Install
Global install:
pi install npm:@codersbrew/pi-tools
Project-local install:
pi install -l npm:@codersbrew/pi-tools
You can also add it manually to pi settings:
{
"packages": ["npm:@codersbrew/pi-tools"]
}
Using the bundled subagent workflows
After installing the package, the prompt templates are available directly in pi:
/plan add Redis caching to the session store
/execute-plan plan/add-redis-caching.md
/implement add Redis caching to the session store
/implement-and-review add input validation to API endpoints
/review src/api/handlers.ts src/api/validation.ts
Tracked plans are written to plan/*.md. New-plan workflows create or reuse a focused git branch before materializing the plan when the checkout is clean enough to do so safely. Tasks are checked off in place, blocked tasks are marked explicitly, and testing/type-checking is required before executors mark implementation tasks complete.
If you want the same tracked-plan output without subagent delegation, use the skill instead:
/skill:plan add Redis caching to the session store
To customize or add agents, create markdown agent files in either:
~/.pi/agent/agents/for your personal defaults.pi/agents/for project-local agents
Package structure
This package uses pi's standard package manifest and currently publishes:
extensions/security.tsextensions/session-breakdown.tsextensions/update-pi.tsextensions/subagent/index.tsskills/github-workflow/SKILL.mdskills/plan/SKILL.mdextensions/subagent/prompts/*.md
The root package.json exposes them through:
{
"pi": {
"extensions": ["./extensions"],
"skills": ["./skills"],
"prompts": ["./extensions/subagent/prompts"]
}
}
Local development
Install dependencies:
npm install
Run local verification:
npm run check
Preview the publish tarball:
npm pack --dry-run
Releasing
This repo includes .github/workflows/release.yml.
Ongoing release flow:
- Update the package version in
package.json - Commit and push the change
- Create and push a matching version tag such as
v0.3.0 - GitHub Actions will run verification, publish to npm, and create a GitHub release
The workflow supports either:
- npm trusted publishing via GitHub OIDC
NPM_TOKENfallback publishing when the secret is present
If you use manual dispatch, ensure the version in package.json has not already been published.
License
MIT