Parcha Skills plugin
A collection of 14 portable skills for long agent runs: work planning, live app QA, search across past sessions and a phone call
- 136 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The skills write reports and edit repository files, `autoqa` runs the app and `review-loop` replies to GitHub threads through your `gh`
- `hands-free` passes agent questions to the external Vapi calling service
Install
- Works in
Manual install
npx skills add Parcha-ai/parcha-skillsFor a single skill add `--skill <name>`, for example `--skill hands-free`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skills write reports and edit repository files, `autoqa` runs the app and `review-loop` replies to GitHub threads through your `gh`
- `hands-free` passes agent questions to the external Vapi calling service
This is third-party code. Review the repository files before installing.
What it does
A repository from the Parcha team with portable skills for Claude Code, Codex and pi. Some drive large work: `parable` plans task batches and hands part of them to cheaper executors, `cascade` keeps a task graph and parallelises execution, and `precap` writes the expected result up front for the agent to check itself against. Others verify output: `autoqa` walks a running app and writes an evidence-backed report, `blast-radius` finds breakage beyond the diff and `verify-release` confirms the intended version works in the target environment. There are session utilities too: `recall` searches past Claude Code and Codex sessions, `recap` reconstructs one session, `tether` keeps Slack threads attached to agents and `hands-free` phones you when the agent needs an answer. Some skills come from `cursor/plugins` with the author and MIT license credited.
Who it is for. Developers who let an agent run on long tasks in Claude Code, Codex or pi and want planning, verification and history search.
Good fit when
- A large project has to be carried through verifiable tasks with parallel work
- Before a release you need a walk through the live app and an evidence report in the PR
- You remember a decision from a past session but not where it was
- An agent is on a long task and needs a phone call to get an answer
Not a fit when
- A one-file task: the full plan and verify cycle adds needless steps
- You run stock pi without a configured CLI executor: `parable` will not work in bare pi
- You do not use GitHub: `review-loop` works only through an authenticated `gh`
Example request
Run autoqa on this repository: start the app, walk the main scenarios and attach an evidence reportLimitations
- Some skills depend on external services: `hands-free` uses a Python and Vapi contract for calls, and `tether` needs a separately installed private runtime.
- In stock pi, `parable` needs a configured CLI-backed executor and `recall` does not index pi's own sessions.
- `blast-radius` runs only on explicit invocation because it sets `disable-model-invocation: true`.
- `review-loop` works only with GitHub through an authenticated `gh`.
- Several skills are adapted from `cursor/plugins`, so upstream updates may diverge from the copies.
How to disable. Delete the relevant skill folders from `~/.claude/skills` or `~/.agents/skills`. If you installed a plugin, remove it with `/plugin uninstall` and the skill name.
FAQ
Which agents are supported?
Claude Code, Codex and pi. The same files are packaged by all three manifests, with no per-agent forks.
How do I install just one skill from the collection?
Run `npx skills add Parcha-ai/parcha-skills --skill autoqa`, substituting the skill you want.
What does autoqa do?
It works out how the repository runs, walks the live app end to end and writes a witnessed report. For UI changes it captures before and after states side by side.
How does recall differ from the built-in agent history?
It indexes past Claude Code and Codex sessions locally and searches them. It runs from pi but does not index pi's own sessions yet.
Who wrote the blast-radius and tdd skills?
They were written by Lauren Tan and live in `cursor/plugins` under the MIT license. The collection adapts them and lists them in its sources table.
Discussion
No questions or reviews yet.
Related
PluginsAI Builder Club Skills
AI Builder Club — Skills
A Claude Code plugin marketplace to make a repo agent-ready and run compounding agent loops
A catalog of nearly two hundred skills: code delivery, review, debugging, AI governance, product work and agent coordination.
Open infrastructure for agents that learn from experience: session traces, a knowledge base and skill packs via CLI, MCP and API
A tool that builds tested skills from a real workflow and governs their lifecycle through your own marketplace
A plugin of seven skills for Claude Code and Codex: prompts, subagents, skill authoring, editing, web search and a browser.
31 skills for Claude Code and Codex: cross-vendor review, task handoff on low quota, a morning brief and a nightly retrospective
README in short
The page opens with a table of skills and compatibility notes for Claude Code, Codex and pi. A sources table for borrowed skills, install commands for the whole collection and for a single skill, and notes on pasteable snippets follow.