Shared Pi coding agent skills for batch processing, automation pipelines, and agent architecture patterns — plus other cross-project, agent-agnostic utilities kept here for sharing.
| Skill | Description |
|---|---|
agent-design-for-batch |
Architectural patterns for building robust batch processing systems with the Pi agent SDK — worker pools, stall detection, graceful shutdown, parallel prompt architectures. |
ai-code-review |
Writing code for LLM maintainability and setting up AI-assisted code review. |
cpp-runtime-audit |
Benchmark, profile, and audit compiled C++ programs (peak RSS, threading, runtime gotchas). |
critical-research |
Develop, deepen, or stress-test research or ideas with critical analysis. |
manim-math-tutor |
一对一数学辅导：解答数学题，生成 HTML 讲解文档和带配音的 Manim 动画视频. |
markitdown |
Convert PDF, Word, and other documents to Markdown. |
mental-models |
Mental models and thinking frameworks (fs.blog collection) for reasoning about problems. |
mean-reversion-testing |
Test for mean reversion in a series (ADF test, stationarity checks). |
visualize |
Add one correct, minimal visual (diagram / geometric picture) when an idea is genuinely clearer as a picture. |
notebook-output-dedup |
Deduplicate repeated output in Jupyter notebooks (e.g. charts rendered twice). |
rust-harden |
Harden Rust code for production. |
rust-runtime-audit |
Benchmark, profile, and audit compiled Rust programs. |
teach |
Teach the user a new skill or concept within the workspace. |
Skills follow the standard Pi skill format (SKILL.md with YAML frontmatter). Install by copying the skill directory into your Pi skills directory; in this workspace they are symlinked from ../skills/ so agents discover them in place.
MIT