skill-draftinglisted

# Skill drafting You manage memsearch's **procedural memory**: remembered work expressed as reusable skills, beside the episodic daily memory files. This manages memsearch's skill candidates, not pi's own skills system. memsearch's background pass (`skills distill`) is unavailable under pi — it needs a memsearch-configured API provider and an upstream `plugins.pi.*` platform that does not exist — so drafting is always done by you, the session agent, in the ways below. Stages: **0** daily memory files → **1** skill candidate (`.memsearch/skill-candidates/`, a git-tracked store that keeps evolving) → **2** installed (a pi skills directory). Candidates are never installed automatically; installing is always a deliberate human step. A request may stop at drafting or review, or continue to installation in the same exchange after explicit approval — match the requested stage. Invoke memsearch exactly as the extension does, through the pinned spec: `uvx --from 'memsearch[onnx]>=0.4.17,<0.5' memsearch …`. ## Intent routing - "make/turn this into a skill", "from what we just did" → **A. Capture now** - "what candidates are there", "review candidates", "install X" → **B. Review & install** - "mine my history", "find recurring workflows" → **C. Mine history** - Unclear or empty → run **B**'s `list`; if empty, offer A or C. ## A. Capture now (0→1) You already have the context, so draft the skill body yourself: markdown without frontmatter, imperative numbered steps for the recurri