agentleFS
Simple-With-UsSimple-With-Us/BotFleet/.claude/skills/deploy-verify/SKILL.md

deploy-verify

Skill4 starsChanged 10 days agoApache-2.0View on GitHub

Verify production after a merge or deploy — per-app health URLs, Coolify vs Vercel vs library-tag, backup continuity, and known failure classes. Use after merging to main, after a Coolify/Vercel deploy, when health looks stale, or when the owner asks if prod is up. Do not treat Coolify API status as truth.

  • Reads credentials

9 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 0 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read /Users/jay/apps/COOLIFY.md before poking the API.

    line 15
  2. Very specific:

    If Infisical still has COOLIFY_API_TOKEN for metrics, it must equal COOLIFY_SERVER_STATS.

    line 30
  3. Very specific:

    Prefer live GET /api/v1/applications (via a helper that reads the token itself, or Coolify MCP) over memorized UUIDs.

    line 44
  4. Very specific:

    Do not interpolate the token into a command you will see.

    line 67
  5. Very specific:

    Do not revert raised tcp_mem sysctl just because a deploy succeeded.

    line 104

“Common” is measured: kept by at least 6 in 10 of the 7 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Deploy verification (MONET)

Shared ~/.claude/skills. Monet, Claude/Fable, and (when active) Renoir all load this directory. Do not treat the word Monet in examples as proof of your seat. Pin AGENT_SEAT / AGENT_TAG from the logged-in account before Slack or board --by:

  • Monet → MONET, Notes Monet, monet/, ~/apps/<app>-monet
  • Claude / Fable → CLAUDE, Notes Claude, claude/, ~/apps/<app>-claude
  • Renoir → RENOIR, Notes Renoir, renoir/, ~/apps/<app>-renoir Cursor, Grok, Grok Bot, Codex, AG, DeepSeek, Kimi, and Fx have their own skill dirs and must not take identity from here.

Read /Users/jay/apps/COOLIFY.md before poking the API. Prefer public health + docker ps on the box over Coolify UI/API status (that field goes stale).

Operator host (live 2026-08-07+): Hetzner NBG1 <PROD_ORIGIN_IP> / fleet-hetzner-nbg1. SSH ssh -i ~/.ssh/hetzner root@<PROD_ORIGIN_IP>. Dashboard https://host.jays.services.

Retired — do not use: <HETZNER_IP_RETIRED>, <ORACLE_IP_RETIRED>, Oracle Tailscale <TAILSCALE_IP_RETIRED>, Coolify UUID <OLD_ST_COOLIFY_APP_UUID>, token name COOLIFY_API_TOKEN on the command line.

Token split

KeyUse
COOLIFY_SERVER_STATSRead-only. App server-stats panels. Infisical metrics token.
COOLIFY_AGENTSFull deploy/admin. Agent ops only. Never store this in Infisical as the app's COOLIFY_API_TOKEN.

Load via env, not argv. Names only: fleet-secrets names. See secret-handoff.

If Infisical still has COOLIFY_API_TOKEN for metrics, it must equal COOLIFY_SERVER_STATS.

Per-app how merge becomes prod

AppProductionMechanismHealth
SThttps://socratictrade.comCoolify auto-deploy on main. UUID <ST_COOLIFY_APP_UUID> (confirm live). Merge == live — do not also click Deploy.curl -fsS -A 'Mozilla/5.0 (compatible; fleet-coolify/1.0)' https://socratictrade.com/api/health
CThttps://congress.tradeCoolify auto-deploy (app/** / services/**). UUID <CT_COOLIFY_APP_UUID>. Worker ship.sh / deploy.yml is leftover.curl -sA 'Mozilla/5.0' https://congress.trade/api/health (bare curl historically 403)
UMhttps://usage.jays.servicesCoolify webhook. UUID <UM_COOLIFY_APP_UUID>. Render and the Oracle auto-deploy timer are retired.curl -fsS https://usage.jays.services/api/health and curl -fsS 'https://usage.jays.services/api/ready?strict=1'
DealDexhttps://dealdex.onlineVercel on merge. Do not Coolify. dealdex.vercel.app is a different Next.js site.curl -sI https://dealdex.online
Personal-Sitehttps://jays.servicesVercel behind Cloudflare. GitHub merge does not auto-publish live. Do not create a second Vercel project.curl -sI https://jays.services
CTSpublished tag vX.Y.ZLibrary. Announce on Slack, then tag. Consumers pin the tag.n/a
AFLhttps://simple-with-us.github.io/AI-Fleet-Coordinator/GitHub Pages digest (case-sensitive path; lowercase 404s today).curl -fsSI https://simple-with-us.github.io/AI-Fleet-Coordinator/ios-versions.json

Prefer live GET /api/v1/applications (via a helper that reads the token itself, or Coolify MCP) over memorized UUIDs.

FLEET-PROTOCOL's "sanctioned deploy path" table still names ST uuid <OLD_ST_COOLIFY_APP_UUID>, CT Worker deploy.yml, and UM Render — stale. Prefer this table + ~/apps/COOLIFY.md (top Hetzner section only; the Oracle body below the cut is historical).

Steady state: ST/CT/UM auto-deploy via Coolify webhook; DealDex via Vercel; AFL digest via Pages; CTS via tag. Announce-then-deploy (~10 min, one deployer) only for a manual Coolify/API click. Do not double-trigger: Coolify cancel is unreliable.

1. App health (always start here)

ST fields are nested under .checks (camelCase):

curl -fsS -A 'Mozilla/5.0 (compatible; fleet-coolify/1.0)' https://socratictrade.com/api/health \
  | jq '.ok, .checks.db, .checks.schedulerAgeSeconds'

Expect ok: true, db ok, scheduler age small if a scheduler exists. Root may 307 to /login.

CT: send a browser User-Agent. A workflow red X on health is often a false Cloudflare challenge, not a failed deploy.

2. Coolify API (optional, often 403)

host.jays.services has sat behind a Cloudflare allowlist. A 403 from your laptop is not a production outage.

Do not interpolate the token into a command you will see. Prefer Coolify MCP or a wrapper that reads COOLIFY_AGENTS itself. Expect status: "finished" only as a hint — then confirm public health and:

ssh -i ~/.ssh/hetzner root@<PROD_ORIGIN_IP> 'docker ps --format "{{.Names}} {{.Status}}"'

Zombie: a deploy stuck in_progress blocks the queue (concurrent_builds serializes). Post #Work-In-Progress with the deployment id if you can see it from the box.

Silent freeze (ST #2545 class): webhook 200 + healthy /api/health on an old sha. Standing watch: ST .github/workflows/deploy-freshness.yml. Do not hand-trigger; inspect the queue.

3. Backups (ST/CT/UM SQLite)

Litestream (or the host /usr/local/sbin/fleet-sqlite-backup.sh) is the DB path. Prefer /api/health storage fields over browsing R2.

ST:

curl -s https://socratictrade.com/api/health \
  | jq '.checks.storage | {litestreamAgeSeconds, litestreamStatus, litestreamState, litestreamDegradedReasons}'

storageDegraded: true or stale/stopped → escalate immediately on Slack + board.

Litestream 0.5.12 is the pin after 0.5.14 leaked TCP sockets (2026-07-10). Do not "upgrade" it casually.

Host scripts (cron): fleet-sqlite-backup.sh, fleet-health-verify.sh, fleet-backup-verify-weekly.sh. Hetzner server backups are the outer layer — not sufficient alone for trading DB RPO.

4. Box checks

ssh -i ~/.ssh/hetzner root@<PROD_ORIGIN_IP>
df -h / | grep -E 'Use%|^/dev'
ps aux | grep -E 'nixpacks|coolify' | grep -v grep

Disk hygiene timer already prunes above high watermarks. Sustained pressure or a stuck build is a wedged queue — Coolify container logs, not a second deploy click.

Do not revert raised tcp_mem sysctl just because a deploy succeeded.

5. After verify

Move the effort row to Deployed only when you say how you verified. closeout skill. Rollback: Coolify previous deployment / Vercel rollback — not Mac pm2, not trading-publish.sh (retired).

Canon

  • /Users/jay/apps/COOLIFY.md
  • /Users/jay/apps/FLEET-PROTOCOL.md — Production deploys; Coolify tokens
  • ST docs/rollouts/2026-08-07-hetzner-fleet-cutover.md
  • ST docs/rollouts/2026-07-10-auto-deploy-on.md (mechanism; UUIDs/IPs in that note may be pre-cutover)
  • Skills: secret-handoff, closeout, land-lane

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.