k1000

Self-evolving agent memory via context information density maximization. Crystallizes completed tasks into reusable Skills (L3), maintains layered memory (L0–L4), and compresses context to preserve decision-relevant signal. Hybrid model: project-scoped (.pi-memory/ in repo) + global (~/.pi/agent/skills/generic-agent/memory/). Project skills shadow global skills. Auto-triggered after reflect_capture to sync learnings, and before tasks to recall relevant skills. Use when building on prior work, iterating on similar tasks, preserving session learnings, or managing long-running sessions.

Maintain durable project documentation and long-term memory using the Archivist pi extension. Use when the user mentions Archivist, knowledge extraction, extracting lessons from papers/books/articles/repos, commit-triggered documentation, Sherpa write-side memory, Obsidian project memory maintenance, catalog.csv navigation, reflection preservation, distillation, documentation audits, safe automations, or installing the Archivist git hook.

Use j01n.me to join an ephemeral collab space with other agents.

Use Sherpa for all context retrieval. Sherpa is the session firewall: it finds, ranks, and distills context so only high-signal output enters the session. This keeps the session clean and small. Use before any non-trivial task. Only bypass Sherpa when you are certain of the exact file path and line number.

Design, create, run, and verify deterministic TDD loop-engineering workflows using the tdd_loop extension. Use when the user asks to build a loop, codify expected coding functionality as tests, create a TDD loop, split a complex feature into test-backed subloops, define completion verification, or run /tdd_loop commands.

Proactively offload lighter bounded side jobs to a lower model through the Bob pi extension so the costly main agent keeps context clean. Use often for work that is useful but not worth main-model context/reasoning: file discovery, first-pass review, summarization, simple investigation, verification advice, or small mechanical edits when explicitly safe. Always provide enough context because Bob may not have the main agent's full context.