agentleFS
NasserAlbusaidiNasserAlbusaidi/ship-gate/skills/deploy-ceremony/SKILL.md

deploy-ceremony

Skill0 starsChanged 4 months agoMITView on GitHub

Deploy the Rihla Firebase backend (Cloud Functions + Firestore rules + indexes) to prod and keep the deploy state honest. Shows the merged-but-undeployed delta first (pending_deploy.sh), runs the commit-bound deploy + prod-state verify, advances the `backend-deployed` source-of-truth tag, then records the deploy in docs/DEPLOY-LEDGER.md and clears stale "⚠️ NOT deployed" flags. Use when asked to "deploy the backend", "run the deploy ceremony", "ship the functions", or after merging a Functions/rules PR.

  • Deletes or force-pushes
  • Commits and pushes

The file

Files
SKILL.md

Deploy ceremony (backend)

The project's recurring failure mode is ghost-debt: Functions/rules merged to main but never deployed, tracked only in scattered ⚠️ NOT deployed memory notes. This ceremony makes deploy state a single queryable fact (the backend-deployed git tag) and a discoverable log (docs/DEPLOY-LEDGER.md), and refuses to let a deploy finish without recording itself.

Scope: this is the backend deploy (Cloud Functions + firestore.rules + firestore.indexes.json + hosting). It is not the client release ceremony (tool/release.sh) and it does not satisfy the release-readiness gate — see the warning in step 5.

Live-users note (supersedes the old "no-clients deploy freely" note)

Real users are LIVE — the no-clients era is over. Client-compat gating applies: a rules/Functions tightening that can deny a shipped client's writes ships client-first or same-release (the ⚠️ old clients PERMISSION_DENIED notes #290/#318/#294 govern again), and backend deploys ride RELEASES, not individual merges — run this ceremony at release time, not after every merge. See CLAUDE.md → Workflow for the governing rule; if [[no-clients-deploy-freely]] still exists in memory, it is stale.

Step 1 — show what's pending

bash tool/pending_deploy.sh rihla-safar
  • Exit 0 (nothing pending) → stop; prod already matches main.
  • Exit 1 → it prints the merged-but-undeployed backend commits + file delta. Read them. State out loud what is about to ship and why.

Step 2 — get onto the exact commit being deployed

The deploy is commit-bound: it refuses unless the worktree is clean and RIHLA_FIREBASE_DEPLOY_APPROVED_SHA equals HEAD.

git checkout main && git pull --ff-only
git status --porcelain   # must be empty

Deploy the tip of main (everything reviewed + merged). If you must deploy a non-tip commit, check it out explicitly and know that anything merged after it will still read as pending afterward.

Step 3 — deploy (commit-bound + prerequisites confirmed)

Prereqs the script enforces: Blaze plan, Functions/Cloud Build/Artifact Registry APIs, App Check enrolled before enforced callables. The --force makes repo config authoritative (removes stale remote indexes/functions).

This needs your Firebase auth and is long-running — run it yourself in the session with ! so its output lands here:

! RIHLA_CONFIRM_FIREBASE_DEPLOY=yes RIHLA_CONFIRM_APP_CHECK_READY=yes RIHLA_FIREBASE_DEPLOY_APPROVED_SHA="$(git rev-parse HEAD)" bash tool/deploy_firebase_backend.sh rihla-safar

On success the script: deploys → runs check_firebase_prod_state.sh → moves + pushes the backend-deployed tag to the deployed SHA. If the tag push warns of failure, push it manually (git push --force origin refs/tags/backend-deployed) or pending_deploy.sh will keep reporting the just-shipped change as pending.

Step 4 — verify it landed

git rev-parse backend-deployed          # should equal the deployed SHA
bash tool/pending_deploy.sh rihla-safar # should now exit 0

A green check_firebase_prod_state.sh inside the deploy is the prod-match proof; do not assert success without it (this repo's signature failure is fabricated done-claims — see CLAUDE.md → Memory).

Step 5 — record + clean up (the honesty step)

  1. Ledger — append a row to docs/DEPLOY-LEDGER.md (newest at bottom): date, deployed SHA, the PRs/issues that shipped (from step 1's delta), and the prod-state result. Open a docs-only PR (main is protected; it's Gate-exempt → /automerge lands it on green readiness).
  2. Memory — flip the ⚠️ NOT deployed flags in the relevant project_*.md memory files (and MEMORY.md index lines) to deployed, with the SHA + date.
  3. PRODUCTION-READINESS — clear the hand-written pending-backend-deploy informational block (the one PR #325 added). Do NOT touch the pinned - [ ] Firebase production state is not aligned with this branch yet. checkbox — test/unit/release_workflow_gate_test.dart keeps that OPEN until the full release ceremony (a recorded prod-state PASS vs the release SHA), which is a higher bar than a backend deploy. Flipping it turns CI red.

Non-negotiables

  • Show pending_deploy.sh BEFORE deploying — never deploy blind.
  • The deploy is the only thing that may advance backend-deployed; never move the tag by hand to paper over a skipped deploy.
  • A deploy that doesn't reach step 5 left ghost-debt — finish the record, or you've recreated the exact problem this ceremony exists to kill.

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.