claude-sdk-rs: Command for Claude Code

Log Work — Append a Log entry, sync status, and regenerate the freshness spine via mev emit-state.

This command is brain.toml-driven and depth-agnostic. It works unchanged whether this repo lives at the brain root, inside a tier sub-brain (core/, portfolio/, side/, client/), or standalone outside any brain — it resolves everything from the manifest at runtime. There are no baked ../ paths and no {{SLUG}} token; if you find either, it is a stale copy.

Variables

$ARGUMENTS — optional free-text explanation of what was done. May be brief or detailed. If provided, it is the primary narrative for the Log entry — do not discard or summarise it down. If omitted, derive the narrative from git history and the task spec.

Execution Model

Run entirely inline. Spawn no subagent. This command authors state (flips closed-block status, appends the Log entry, surgically updates status.md, offers to record a decision) and then shells out to mev emit-state --write, which is what actually regenerates the freshness spine (the watermarked caches + generated rollups that mev validate-brain --sync checks) — so favour accuracy over speed in the authored parts. A subagent cold-starts and has to reconstruct this session's narrative from git log alone; the main agent already holds the session context needed for an accurate log.md entry, and a round trip through a subagent adds latency without adding value here.

Instructions

Step 0 — Resolve the manifest (this is what replaced the baked paths)

1. Find the brain root. From the current working directory, walk up parent by parent looking for a brain.toml file (its first line begins # brain.toml). The directory containing it is BRAIN_ROOT.
   • If no brain.toml is found, this is a standalone repo. Skip the brain-sync Step 3 (mev emit-state --write has nothing to resolve without a brain.toml): do only Steps 1–2 (local log.md + planning/status.md), then go to Step 4 and say so in the report.
2. Identify this repo in the manifest. Compute this repo's path relative to BRAIN_ROOT (. if you are at the root). Read brain.toml and find the [[repos]] entry whose repo_path equals that relative path. From it read: slug, tier, status_file, cache_doc, heading.
   • If no entry matches (a repo not yet registered), STOP and report it — do not guess a cache path. (Registering a new repo is /new-project / /generate-sub-brain's job.)
   • All status_file / cache_doc paths in the manifest are relative to BRAIN_ROOT — resolve them against BRAIN_ROOT, never against the current directory.