Operations

PROMPT 1 of 2 — learn what this module teaches. You write nothing anywhere in this prompt. The only commands you run are the read-only curls below.

ASK EVERY QUESTION WITH the `AskUserQuestion` tool, with a RECOMMENDED option first. Not prose.

Fetch and read the whole page:

  curl -s https://docs.utopiamodels.ai/docs/operating/operations.md

If the fetch fails, say what came back and stop.

1 — EXPLAIN IT BACK, UNDER THESE HEADINGS

  EVERY COMMAND NAMES ITS BOX  The audit that produced this rule — how many self-checks failed from the wrong box,
                               and how many of those passed on the right one. Why "accurate and location-blind" is
                               still broken.
  IS IT ALIVE?                 Why `getent passwd` is not a liveness check, what the real probe exercises, and what a
                               403 means about the account versus the box.
  A CLEAN --failed             The two traps that produce a clean bill of health on a broken box.
  DELETING A UNIT              Why removing it from the repository does not stop it, and the two halves.
  A 200 IS NOT PUBLIC          What `redirects=0` and the final URL add that a status code cannot, and the difference
                               between DEPLOYMENT_NOT_FOUND and a 530.
  SHIPPING                     Why a merge is never a human gate, the three failure branches, and why a bare
                               `git push --force` is banned.
  VERIFYING A CLAIM            The four habits, in order.

Quote the page where you lean on it.

2 — RUN THE PUBLIC-SURFACE PROBE, LIVE

Against the page you just read:

  curl -s -o /dev/null -w '%{http_code} redirects=%{num_redirects} final=%{url_effective}\n' -L https://docs.utopiamodels.ai/docs/operating/operations
  curl -sI https://docs.utopiamodels.ai/ | grep -iE 'server|x-vercel|cf-ray|x-matched-path'

Paste the raw output. From it alone: is this page public, which edge served it, and what would the output look like if an SSO gate sat in front of it? Then apply the page's fourth habit — PREFER THE NEGATIVE CONTROL — and run the first command against a URL you expect to fail (for example a path that does not exist on the same host). Show that the check can go red.

3 — QUIZ ME

Three questions with the `AskUserQuestion` tool, one at a time, four options each, correct answer not always first. Draw from: what `resets=unknown` means, why a secret is passed on STDIN, what an enabled timer with a `not-found` service does, and why `--force-with-lease` names the branch. After each answer, quote the line that settles it.

If `AskUserQuestion` is unavailable in this session, ask the same questions as plain numbered text in one message instead.

Then say exactly this:

  "Module learned. Paste prompt 2 and I will audit your own runbooks against it."

Then stop.

PROMPT 2 of 2 — integrate this module into my own setup. Survey read-only first; propose; write only what I confirm.

HOW LONG THIS TAKES: a multi-step audit (7 steps, about 4 questions, writes only what you approve) — expect roughly 10–25 minutes across several turns. Running headless with a timeout? Budget for that, or start with a shorter prompt first. Cut off partway? Paste this prompt again — it resumes from its checklist rather than starting over.

ASK EVERY QUESTION WITH the `AskUserQuestion` tool, each with a RECOMMENDED option first and one line of why.

Reference, read it first:

  curl -s https://docs.utopiamodels.ai/docs/operating/operations.md

FIRST — WRITE YOUR CHECKLIST TO A FILE

RESUMING? Before you write anything, check whether ~/OPERATIONS-INTEGRATE-CHECKLIST.md already exists. If it does, this prompt was interrupted: read it, tell me which steps are already ticked, and continue from the first unchecked step. Do not overwrite it, do not redo a ticked step, and do not re-ask a question whose answer it already records — the scope answer under ## Rules included, if it's there. If ## Rules has no Scope field, this checklist predates the scope question: say so plainly in your first output line and ask it now, before anything else. Only if the file does not exist, write it fresh as below.

Write ~/OPERATIONS-INTEGRATE-CHECKLIST.md and work from it:

  # Operations module — integrate
  - [ ] 1. Find my runbooks
  - [ ] 2. Grade every command for its box
  - [ ] 3. Grade every health check for liveness vs existence
  - [ ] 4. Find scheduled jobs that cannot fail visibly
  - [ ] 5. Find secrets in argv
  - [ ] 6. Propose the rewrites
  - [ ] 7. Write only what I confirmed
  ## Rules
  - Read only until step 7. No stopping, starting, enabling or deleting any unit, timer or cron entry.
  - I report that a credential is SET, never its value.
  - Every finding carries the file and line, or the command and its raw output.

SCOPE — ASK ME BEFORE YOU SURVEY ANYTHING OF MINE

Before step 1, ask me this with the `AskUserQuestion` tool, with both options exactly as written:

  "Should I treat any path as off-limits, or do you want me to have full access to survey and propose changes anywhere in this environment?"

  A — Exclude .env*, credentials and CI/deploy config (.github/workflows/, deploy keys) from anything I read or propose touching.   [RECOMMENDED for most readers]
  B — No exclusions. Full access to survey and propose anywhere, including secrets and CI files.

This is my preference, not a policy: some people want their agent to reach everything, some do not, and neither answer is the correct one. Never assume it and never pick for me. My answer governs every read and every proposal for the rest of this prompt. Write my answer under ## Rules in the checklist as "Scope: A" or "Scope: B", so it survives compaction. It governs the SURVEY and PROPOSE steps only — nothing gets written until step 7, and there only the runbook diffs I confirm.

Before you ask, decide whether an answer can reach me at all. It cannot in a one-shot non-interactive run — `claude -p`, `codex exec`, `gemini -p`, a scripted or CI invocation — and you can tell without guessing: the `AskUserQuestion` tool is missing from your tool list there, while an interactive session has it. It also cannot if the question comes back immediately with no real answer. In either case do not ask, do not wait and do not re-ask: a question asked as plain text in a one-shot run ends the run with nothing done. This is the one case where you pick for me: treat A as my answer, record and hold it exactly as above, and make the first line of your output exactly: "No interactive scope answer available — defaulted to A (excluded .env*, credentials and CI/deploy config). To choose B, re-run this prompt in an interactive session and answer the scope question."

Only if the `AskUserQuestion` tool is in your tool list but fails when you call it, ask the same question as plain numbered text in one message instead.

1 — FIND MY RUNBOOKS

Every document or script a human or agent follows to operate something: RUNBOOK / OPERATIONS / OPS / DEPLOY / ONCALL files, a docs directory, a Makefile or justfile with ops targets, the agent instructions file (CLAUDE.md and anything it imports). List them with paths.

2 — DOES EVERY COMMAND NAME ITS BOX?

For every shell command in those runbooks: does it say which machine it runs on? Flag each that does not and that depends on location — a localhost port, a local path, a tailnet or private IP, an SSH alias. For every SSH host alias, check whether it resolves from THIS machine (`ssh -G <alias> | grep -E '^(hostname|user) '`); an alias that only expands on one box is the page's `Name or service not known` failure.

3 — LIVENESS OR EXISTENCE?

For every health or status check: does it EXERCISE the thing (make it do one real unit of work) or only prove it EXISTS (a user, a file, a process, a 200)? Quote each. For each HTTP check, does it assert redirects and the final URL, or only the code?

4 — WHAT CANNOT FAIL VISIBLY?

If systemd is present: `systemctl --user list-timers --all` and `systemctl list-timers --all`; for each timer, check its service with `systemctl --user status <svc> --no-pager 2>&1 | head -3` and flag any that is not-found. Then `crontab -l` — a second scheduling layer a clean `--failed` never sees. Compare both lists against what the repository says should be scheduled: a unit the repo deleted that still runs here is the page's two-half deletion, only half done.

5 — SECRETS IN ARGV

Search the runbooks and scripts for a credential passed as a command-line argument (`--token`, `-p`, `password=`, an `Authorization:` header built from a literal). Report file and line — never the value.

6 — PROPOSE

A table per category: finding · file:line · the rewrite. Rewrites follow the page: a `# on <box>` comment on every location-dependent command; a real-work probe beside every existence check; `redirects=0` asserted on HTTP checks; secrets on STDIN; and a negative control for each check so it can be seen to go red.

Ask me with the `AskUserQuestion` tool which category to apply first. RECOMMEND the box labels — mechanical, zero-risk, and the page measured them manufacturing twenty-six false failures.

7 — WRITE ONLY WHAT I CONFIRMED

Edit only the runbooks I confirmed, showing me each diff before applying it. Never touch a live timer, unit or crontab — for orphans, print the two-half removal commands for me to run. If something does not work, say so and stop.

When the writes are done, tell me in one line, filled in with the real paths:

  "To undo, delete `~/OPERATIONS-INTEGRATE-CHECKLIST.md` and run `git checkout -- <each runbook you edited, by path>` (or `git revert <sha>` once committed) — nothing else on your system was touched."

FINISH

Every file changed with its path, the count of findings per category before and after, and the one check you added a negative control for, with the output showing it go red.