baton-inventorylisted
Install: claude install-skill StanimirTenev/baton
# Baton inventory — map the work that already exists
Baton starts empty, but the machine does not. This walks the machine once, finds the work
that was done before Baton existed, and turns it into task folders — **after the human
confirms**. Nothing is moved, and no folder is created before that confirmation.
Task root and logbook name: the session-start board names the task root in its first line
("Baton — the tasks in …"); installed as a plugin, the rules that come with it name both. With
the install script they are also in `~/.claude/baton/hooks/baton.local.json` (`home`,
`logbook`; defaults `~/tasks`, `LOGBOOK.md`).
## 0. Start from the survey
Run the read-only survey first — `python3 <repo>/tools/baton_survey.py`, or
`python3 "${CLAUDE_PLUGIN_ROOT}/tools/baton_survey.py"` in plugin mode. It names every memory
folder, the places memory and instructions point at (other drives included), dated logbook
files and folders that already keep a logbook. Hand those to the agents below: agent 3 reads
**every** memory folder it lists, agent 4 also covers its places outside the home directory,
and the dated logbook files are the record to reconstruct from, grouped by their topic prefix.
## 1. Scan — four read-only agents in parallel
Launch four read-only (Explore) agents in one message. Each returns one table row per
work item: `name (kebab) | paths | last activity | evidence | proposed state | next step | confidence`.
Tell every agent: skip `node_modules`, virtualenvs, `.git`