lever-status

lever-status

Three modes against .agents/levers/: list, detail, cancel. List and detail are read-only; cancel is the only write path. None of them advance the chain — that's /lever <id>'s job.

State model recap

lever.yaml.step ∈ {plan, do, check, act, done, cancel} is the single state pointer. pause ∈ {ask, blocked} appears only when halted at step waiting for the user, and only ever at step: plan or step: do. The pair (step, pause) is the full state.

1. Pre-flight

1. Confirm .agents/levers/ exists. If missing, stop with a chat sentence pointing to /lever-init.
2. Branch on the argument:
   • No argument → list mode (§2).
   • <id-or-slug> alone → detail mode (§3).
   • <id-or-slug> cancel [<reason>] → cancel mode (§4).

2. List mode (no argument)

1. Read every .agents/levers/*/lever.yaml. Skip directories without one and note them under a Skipped: line in the chat reply.

2. For each lever extract: lever_id, slug, step, pause (optional), criteria pass count, updated_at.

3. Print as a plain-text table sorted by lever_id (no markdown table syntax — chat is monospace). State is <step> or <step> · paused: <reason>:

   ID  Slug                     State                  Pass     Updated
    1  add_password_sign_in     check                  3/4      2026-05-03 09:55
    2  fix_dropdown_ios         plan · paused: ask     0/0      2026-05-02 14:12
    3  refactor_api_routes      done                   8/8      2026-05-01 18:30
   

4. Trim slug to 24 chars with … ellipsis if longer. Trim Updated to YYYY-MM-DD HH:MM.

5. End with a one-line summary (e.g., "3 levers · 1 done · 1 awaiting answer · 1 ready to advance.").

3. Detail mode (<id-or-slug>)

1. Resolve <id-or-slug> → .agents/levers/<id>-<slug>/:
   • Pure integer → match the leading <id>.
   • String → match a slug fragment; on multi-match pick the lowest <id> and note in the chat reply.
   • Missing → stop with a chat sentence; recommend /lever-status to list.