A collection of custom skills and extensions for the pi coding agent, designed to run on Termux (Android).
Bridges the gap between a terminal-based AI coding assistant and the mobile environment — enabling document management, web research, phone file navigation, and enhanced text input on Android.
| What | Extensions | Skills |
|---|---|---|
| Document organization — AI generates clean filenames for downloaded PDFs/DOCX with a review UI | docorg.ts |
docorg |
| Document ingestion — Parse PDFs, DOCX, XLSX, images into LLM context | — | liteparse-ingest |
| Phone storage nav — find files by type instantly via pre-built index | — | storage-map |
| Android input — hardware keyboard + on-screen dialog + speech-to-text | termux-input.ts |
— |
| Collapsed output — clean tool rendering, press Ctrl+O to expand | collapsed-tools.ts |
— |
| Web search & research — Tavily CLI integration for search, extract, crawl, deep research | — | 8 tavily-* skills |
git clone https://github.com/a2ajinkya/phone-pi.git
cd phone-pi
pi --agent-dir .Or symlink the pieces you want into ~/.pi/agent/.
Dependencies vary by skill — check individual skill files for details.
| Skill | Description |
|---|---|
docorg |
Organize document downloads with AI-generated filenames and batch review. |
liteparse-ingest |
Ingest PDFs, DOCX, XLSX, images via LiteParse. |
storage-map |
Precise phone storage navigation using a pre-built index. |
tavily-search |
LLM-optimized web search via Tavily CLI. |
tavily-extract |
Extract clean markdown from specific URLs. |
tavily-crawl |
Crawl websites and save pages as local markdown. |
tavily-map |
Discover all URLs on a website without extracting content. |
tavily-research |
Deep multi-source AI research with citations. |
tavily-dynamic-search |
Programmatic web search with context isolation. |
tavily-cli |
Unified Tavily CLI integration (search, extract, crawl, research). |
tavily-best-practices |
Production-ready Tavily integration patterns. |
MIT — use freely, adapt as you like.