I build tools for the Pi coding agent and beyond.

Harness minimalism is a core principle I rely on to guide my development and agent workflows:

• ✅ Keep always-on context lean and bring in task-specific information only when needed.
• ✅ Start with a minimal baseline: clear goals, essential context, success criteria, and firm safety, authorization, and data protection boundaries.
• ✅ Let the model choose its approach; add procedures only to address recurring failures, and retain them only when their benefits justify their costs.
• ✅ Remember, MANAGING CONTEXT is all you need! Leave the rest to the model. 🤓

• pi-agent-kit — Curated skills and extensions for Pi.
• pi-subagent — Bounded, read-only subagent investigations for Pi.

• pi-jev — Experimental, consent-gated Jev routing and public-passage reranking for Pi.