agentleFS
avelikiyavelikiy/great_cto/skills/deploy-landed/SKILL.md

deploy-landed

Skill112 starsChanged 2 months agoMITView on GitHub

Prove a deploy landed — the served revision is the commit you meant, its env is present, and the product does its job — instead of trusting the deploy command's exit code. Also catches server drift before a pull or deploy. Use after any deploy, rollout, restart or server-side fix ("задеплой", "выкати", "проверь на проде", "почему не обновилось"). Loaded by devops, l3-support, infra-provisioner.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before pulling or building on a host someone can reach by hand:

    line 34
  2. Very specific:

    After --set-secrets/--set-env-vars, check that the ones you did not name are still there: those flags replace the whole set.

    line 56

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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-landed — a deploy is done when the running thing is the thing you meant

A deploy command's exit code says the command finished. It does not say the new code is serving, that its configuration arrived, or that the product works. Every trap below produced a green exit and a broken or stale product on a real project:

TrapWhat happened
git pull failed, the build went onThe build succeeded from the old code and reported success
nohup … &Exit 0 twice for a process that died at once
docker exec into a containerIt does not inherit the entrypoint's env: two "config is missing" readings were false
sed -i / docker cp on a bind-mounted fileReplaced the inode; the container kept reading the old file
A fix applied only on the serverLost on the next deploy — it was never in git
wrangler / Pagesfunctions/ not deployed, _headers ignored, five deploys to notice
Prod behind SSOAnswers 401/302 to every path, including ones that do not exist — proves nothing about a route
/health greenFor six hours while the engine made no trades

Before: is the server what git says it is?

Before pulling or building on a host someone can reach by hand:

git -C /srv/app status --porcelain        # anything here is a patch that is not in git
git -C /srv/app log -1 --format=%H        # what the host thinks it runs

Anything uncommitted on the host is either committed first or reported as drift — never overwritten silently, and never left as the only copy of a fix.

After: four checks, each with its evidence

  1. Revision. What is serving equals the commit you meant.
    • Cloud Run / k8s: the live revision's image tag or digest, then git merge-base --is-ancestor <commit> <image-tag-sha>.
    • Pages / Workers / Vercel: the deployment id and its commit.
    • A plain host: the running process's start time is after the deploy, and the file it serves has the new content (grep for a string the change introduced).
    • A /version endpoint, when the app has one, beats all of the above.
  2. Configuration. The variables and secrets the new code reads are present in the running process — count them against the list the code expects. In a container, read them from PID 1 (cat /proc/1/environ | tr '\0' '\n'), not from docker exec env. After --set-secrets/--set-env-vars, check that the ones you did not name are still there: those flags replace the whole set.
  3. Function. One request that does the product's job and returns something you can compare to a known value — an order created, a row written, a trade placed on paper. A health endpoint proves the process is up, nothing more.
  4. Startup errors. Logs for the new revision since it started, at WARNING and above. Zero lines is a finding only if you saw the logs flowing at all.

Report

End the deploy report with this block. A check you could not run is NOT CHECKED with the reason — never omitted, never PASS.

DEPLOY-LANDED <service> @ <commit>
  revision : PASS — serving <revision/deployment id> built from <sha>
  config   : PASS — 14/14 expected vars present in PID 1
  function : PASS — POST /orders → 201, row id 8812 visible in DB
  startup  : PASS — 0 WARNING+ lines since 12:04:10

A FAIL in any line means the deploy is not done, whatever the command printed.

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.