Finding the usage record and the rein command

rein keeps these two out of the session start on purpose: most sessions never need them, so carrying them in every context costs more than looking them up on the one turn that does. Both vary per session and per machine, so take them from the commands below, never from memory or a guess.

In a session rein launched

Run this as-is (it prints both, or stops with the reason):

: "${CLAUDE_CODE_SESSION_ID:?Claude Code did not pass this session its ID -- stop and ask the user for it}"
: "${REIN_MANAGED_CWD:?rein did not launch this session -- use the steps for a session rein did not launch}"
rein_cmd="$(command -v rein)" || { printf 'rein command: not on PATH (report this to the user)\n'; exit 1; }
printf 'rein command: %s\n' "$rein_cmd"
usage_dir="$(rein --cwd "$REIN_MANAGED_CWD" --config "$REIN_MANAGED_CONFIG_FILE" config get usage_state_dir)" || exit 1
printf 'usage record: %s/%s.json\n' "$usage_dir" "$CLAUDE_CODE_SESSION_ID"

• The usage record is <usage_state_dir>/<session ID>.json -- the same file rein's hooks read. cat it: used_percentage is the figure, and at is when the status line last wrote it (an at far behind the clock means the number is old -- rein's hooks say so on their own once it passes usage_stale_sec).
• The location comes from the lineage's own config. A session rein launched carries the lineage's config file in REIN_MANAGED_CONFIG_FILE, and that is what the hooks read; running rein config get without --config would read the default user config instead, which is a different file for a lineage started with --root or --config.
• The session ID comes from CLAUDE_CODE_SESSION_ID. This rests on observation, not on a documented guarantee: Claude Code sets it in the shell its Bash tool runs, with the same value hooks receive as session_id. A subagent's shell carries its parent's ID, so a subagent reads the primary session's record. If it is unset, stop and ask -- never put another session's ID in its place (another session's record reads like a real answer).
• The rein command is whatever rein resolves to on PATH -- the same command a handover request runs (see /rein:request). If it is not on PATH, the installation is incomplete: hand that to the user rather than substituting another path.

In a session rein did not launch

Without the variables above, name the project yourself:

command -v rein
rein --cwd <absolute path to the target project> config get usage_state_dir

The record is <printed location>/<this session's ID>.json. If you don't know this session's ID, stop and ask.

When a step fails

A non-zero exit means the location could not be resolved -- the reason is on stderr (for example, a project config nobody has decided on yet: rein config get refuses to run until it is allowed or denied). Report that reason to the user; don't fall back to a default path or a file you found by listing the directory.