a watcher needs a baseline, and the deploy procedure needs to be findable

Two gaps found by testing the actual goal — "I say watch master for a new tag, and it works without me
listing the caveats" — rather than by rereading what was written.

THE BASELINE. The commander skill had "what must never happen twice", which is a CLAIM: do not act twice on
the same item. The owner's constraint is different and was missing entirely: do not act on items that already
existed when you were armed. "Watch for a new tag" is not a condition — a watcher armed today sees every
historical tag and, unless something says otherwise, treats all of them as news and acts on each in turn.
Nobody says "ignore the ones already there" because to a person it is obviously implied. It is not implied to
a shell loop.

So the skill now names the two honest kinds of baseline — a record of what has been handled, which survives a
restart, or a snapshot taken at arming, which does not — and requires the loop to say which it is using. On
this machine the lock files made the omission harmless. On a fresh one, a new commander would have deployed
every tag in the history, one at a time.

FINDABILITY. The deploy procedure was reachable only through a sentence at the end of the commander skill, so
an agent asked to deploy would have to open one skill and read to its last line to learn the other existed.
It has a description of its own now, which is the entire interface: the agent sees only that before deciding
to open anything.

The new skill does NOT copy the procedure. It names `scripts/deploy-claim.sh`, `docs/servers.md` and the plan
format, which live with the code and change with it — a second copy would be right the day it was written and
wrong the day somebody edits the first, and the copy an agent happened to read would be the stale one. One
hop costs a Read; a wrong procedure costs a fleet.

What it does carry is the fleet-specific part that would not belong in a procedure: poll a local remote and
why (a tag pushed at 21:47 noticed at 23:07 through the public one, twenty seconds through the local),
pm_cwd rather than a hard-coded path, the per-host login, plugins:update restarting sidecars on its own, and
that a skipped host is either behind or current — different things, and only one of them asks the owner for
anything.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

skills/commander/SKILL.md

@@ -53,6 +53,22 @@ if it is money, or a duplicate message, or a half-applied change, the action mus
 starting and refuse a second attempt. One attempt per unit of work, enforced by something the loop cannot
 undo — a file created with `mkdir`, not a flag it sets itself.
 
+**What does "new" mean — and this one you must supply, not ask for.** "Watch for a new tag" is not a
+condition. A watcher armed today sees everything that already exists, and unless something says otherwise it
+will treat all of it as news and act on every item in turn. Nobody says "and ignore the ones already there",
+because to a person that is obviously implied. It is not implied to a shell loop.
+
+So a watcher needs a **baseline**, and there are only two honest kinds:
+
+- **A record of what has been handled.** Best where acting is expensive or must not repeat: the loop skips
+  anything with a record and creates one before acting. This also survives a restart, which a remembered
+  baseline does not.
+- **A snapshot taken at arming.** Fine where the history is genuinely irrelevant — "anything that lands from
+  now on". Cheap, and it forgets everything the moment the loop is restarted, so say that out loud.
+
+Choose one and say which, in the condition itself. A watcher with neither looks identical to a working one
+until the moment it fires on a backlog.
+
 **What does "done" look like to you?** A report is worth more than a success flag, and the parts worth
 reading are the ones that are hard to reconstruct: what it skipped, what it could not verify, what it left
 outstanding. Ask for those explicitly or you will get a summary of what went right.

skills/deploy/SKILL.md

@@ -0,0 +1,77 @@
+---
+name: Deploying Officer
+description: >-
+  Use when asked to deploy, release or roll out Officer, to watch for a release and deploy it, or when a
+  deploy has gone wrong — a host behind, a tag not noticed, a rollout that stopped halfway. Points at the
+  procedure, and carries the things about THIS fleet that are not in it.
+version: 1
+---
+
+# The procedure is a file, and it is not here
+
+Read it:
+
+```
+~/officer-dev/platform/prompts/deploy-commander.md
+```
+
+**Not copied into this skill, deliberately.** It names `scripts/deploy-claim.sh`, `docs/servers.md` and the
+plan format, all of which live in that repository and change with the code. A second copy here would be
+right on the day it was written and wrong on the day somebody edits the first one — and the copy an agent
+happens to read would be the stale one. One hop costs a `Read`; a wrong procedure costs a fleet.
+
+That file carries the whole rollout: what wakes you, what a release is, the per-host order, what to refuse,
+and why each rule exists. Everything below is what is true about this fleet and would not belong in it.
+
+# What a release is here
+
+A **tag**, `deploy/<version>`, version `yyyy.mm.dd.NN`. Nothing else triggers a deploy — not a commit
+message, not a branch, not a push. A tag cannot be typed into prose by accident, which a code word can, and
+`git ls-remote --tags` answers "is there one" in a single call without fetching.
+
+`package.json` must agree with the tag, and `deploy/<version>.md` — the plan — must exist. If either is
+untrue, **stop and say so**. A release with no plan is not one you can guess at.
+
+# One attempt per release, and it is not yours to enforce
+
+```bash
+scripts/deploy-claim.sh claim <version>    # exit 0 = yours, exit 1 = do not proceed
+scripts/deploy-claim.sh done  <version>    # after it succeeded
+```
+
+Exit 1 means somebody already attempted this release, and that answer never changes back. **A failed release
+is not retried, it is looked at.** Clearing a claim is a human act.
+
+This is a script rather than a rule you follow because the watcher loop is rewritten from scratch every time
+one arms, and the wrong implementation is the one that looks right: test the `.done` file, which exists only
+on SUCCESS. Write that and a failing release spawns a fresh agent every poll, all night, each one costing
+money. That is not hypothetical.
+
+# Fleet facts that cost time to learn
+
+- **Poll a LOCAL remote.** `origin` on `officer-dev` points at the local gitea. The public hostname resolves
+  through a reverse proxy where every connection shares one source IP, and rate limiting there refuses SSH
+  for long stretches — presenting as `Connection closed by <ip>` and "make sure you have the correct access
+  rights", which is not an auth problem. Cost: a tag pushed at 21:47 was noticed at 23:07. With the local
+  remote, twenty seconds.
+- **Never hard-code a checkout path.** Take it from PM2 — the `officer` process's `pm_cwd`. One host carries a
+  second, stale checkout, and pulling that one succeeds while changing nothing.
+- **The login user is not always the same.** It is per host, in `docs/servers.md`. A wrong username presents
+  as `Permission denied (publickey)`, which reads as a broken key.
+- **`bun run plugins:update` restarts every plugin's sidecar itself.** A report listing only `officer` under
+  "restarted" understates what moved by an order of magnitude.
+- **Skip the host you are running on, entirely.** Restarting `officer` there ends the rollout mid-report.
+  Name it as skipped, and say whether it is behind or already current — those are different, and only one of
+  them asks the owner for anything.
+- **A dirty tree means TRACKED changes.** `git status --porcelain` counts untracked files; a stray script a
+  fast-forward could never touch is not a reason to skip a host. Use `--untracked-files=no`.
+
+# The report is an artefact, the lock is the record
+
+Write the local record FIRST, then commit and push the report. If the push fails, say so and stop: the
+deploy succeeded and its paperwork is pending. Retrying a push into a loop, or treating a failed push as a
+failed deploy, are both worse than a missing document.
+
+Say what you did NOT do. A host skipped, a step you could not run, a verification you had to perform
+differently — those are the parts of a report anybody actually needs, and the only ones that are hard to
+reconstruct afterwards.