Update

arlo:update — re-sync arlo with the ground it stands on

arlo's whole reason to exist is that stored command knowledge rots: a hand-maintained table said --parent for four months after the real flag became --parent-id, and everyone who trusted it created orphans. A .arlo/cards.json left alone becomes exactly that table. So cards are a regenerated projection of live ground truth, never stored truth — and this skill is how you regenerate them. Run it periodically, and after anything changes the project's commands.

Update reconciles this project against the arlo you already have installed. It does not fetch a newer arlo — that is /arlo:upgrade (pull the latest arlo from upstream, then run this to reconcile). Keep the two axes distinct: upgrade changes arlo; update re-reads your project.

REFERENCE.md (bundled beside this file) carries the full harvest rules (§2) and gradient.

Steps

1. Sync the project's core to the installed arlo. Copy the core the lights-out arlo CLI runs (under arlo's home) from the currently installed plugin, so the project runs the arlo you have. (If you want a newer arlo — new harvest rules, a new rung, a sharper ranking — run /arlo:upgrade first; it fetches the latest arlo, then runs this step.)

2. Re-harvest from the current live sources. Re-run the harvest named by the project's card spec against the sources as they are right now — the same generation as setup, run fresh — into a new cards set:

   python3 -m arlo.cards <spec.json> --root . --out .arlo/cards.json.new
   

   Apply §2's completeness rules again; the surface may have grown or shrunk since setup.

3. Diff, and report the drift out loud. Compare the new cards against the cached .arlo/cards.json and tell the operator what changed:

   • added — a real command that now exists and had no card,
   • removed — a card whose source no longer exists (a moved or deleted script/target),
   • changed — same intent, different command skeleton or purpose (a renamed flag, a rewritten usage). This is the --parent → --parent-id case; surface it explicitly, it is the single most important thing update finds. If a source named in the spec has vanished, say so — do not silently drop it.

4. Promote and re-stamp. Replace .arlo/cards.json with the new set and stamp it with the current git SHA / mtime. Preserve .arlo/config.json — the model tier is the operator's decision, not something to re-resolve unless they ask. Regenerate .arlo/LIGHTS-OUT.md from the refreshed cards.

5. Re-verify. Run a couple of real intents through arlo <words> and confirm each answer is still a real, grounded command.

Update is idempotent and safe to run often — on an unchanged project it re-harvests to the same cards and reports "no drift." The cost of running it too often is nil; the cost of not running it is arlo becoming the stale table it was built to replace.