Analysis updated 2026-05-18
Add a context-checkpoint tool to a Pi agent session so progress can be summarized mid-task.
Install a semantic code-search subagent instead of relying on exact-text search.
Pull in ready-made skills for GitHub issue and pull request workflows.
Rewrite AI-generated text to sound less like corporate marketing copy.
| igorwarzocha/howaboua-pi-stuff | apex-quant-systems/polymarket-weather-trading-bot | marketpulselab/solana-rug-checker-bot | |
|---|---|---|---|
| Stars | 50 | 50 | 50 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | easy | hard | moderate |
| Complexity | 2/5 | 4/5 | 2/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
This is a personal collection of add-ons for Pi, an AI coding agent platform. The author published them as separate npm packages so anyone can install the full set or pick only the pieces they need. Each package is small and focused on one thing rather than trying to be a framework. The packages fall into two categories. Extensions add new tools to the agent: one lets the agent adjust how much it reasons depending on the task, another adds session markers so you can set a checkpoint mid-session and summarize progress before continuing, a third adds a read-only subagent that searches code by meaning rather than exact text, and several others handle things like side-session questions, local memory, and code review passes. Skills are reusable instruction sets the agent follows when you invoke them: there are skills for GitHub issue and pull request workflows, for refactoring agent-generated code into cleaner shapes, for rewriting text to sound less like a corporate brochure, and for browser control via Chrome's debugging protocol. The intended workflow is fairly straightforward. You start a Pi session, let the agent explore the codebase, pull in open GitHub issues, set a context marker, work on one feature or fix, run a review subagent to check the work, then open a pull request. After larger changes you run a hardening pass to tidy the output. The README is honest that this is not about agent swarms or automation magic, it is about keeping sessions focused with clear context boundaries and a handful of tools that reduce repetitive friction. A few packages are intentionally left out of the bundles because they depend on a specific model, desktop setup, or personal taste. The author lists those separately with links to their original sources. All current packages are MIT-licensed. The repo also points to a few external skills the author uses but does not bundle, covering things like frontend polish and UI generation.
A collection of small npm add-on packages, extensions and skills, for the Pi AI coding agent, covering things like context checkpoints, code search, and GitHub workflows.
Mainly TypeScript. The stack also includes TypeScript, npm, Pi agent.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.