garrytan / garrytan/gstack

[NEW] Silent deploy failures section for review/checklist.md

Open Beginner friendly
#2,089 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
TypeScript
Stars
133k
Forks
19.9k
Avg merge
18h 46m
Merged PRs (30d)
26

Description

## What's missing from the review checklist

The current `review/checklist.md` is strong on code-diff issues (SQL safety, race conditions, LLM trust boundary). It doesn't cover the class of failures that pass every local test and surface in production — specifically around the deploy pipeline itself, not the code.

## Proposed addition: Silent deploy failures

A new section at the end of the checklist (or a parallel `deploy-checklist.md`) covering:

**Build cache serving a prior artifact**
The build command reports success and the artifact hash looks right — but the cache layer served a build from before your latest change. The deploy exits 0 with the wrong code.

**Release pointer not bumped**
The deploy command targets a tag, manifest version, or image label that wasn't updated. It deploys the previous build, exits 0, and the health endpoint returns the old version. "Deploy succeeded" and "new code is live" are different claims.

**Migration not applied in prod**
Auto-runs in dev and staging. Requires a manual apply step in prod (or a specific deploy flag). The new code ships, hits the old schema, and fails at runtime.

**Feature gate inverted between environments**
Flag is on in staging, off in prod. The test that passed is testing the opposite of what prod will run.

**Staged rollout masking**
Health check passes on the canary slice. The full rollout has a different failure mode that doesn't surface until traffic shifts.

## Why this belongs in gstack's checklist

These aren't caught by code review — they're post-merge, deploy-time failures. The `/review` skill runs before merging; a "silent deploy failures" scan should run before or after `/ship`. The operating-kit's [field-notes.md](https://github.com/Sharrmavishal/operating-kit/blob/main/docs/claude/field-notes.md) covers this class of failure in detail if useful as reference.

Happy to draft the actual checklist section if this direction is welcome.

Contributor guide

Open the contributing guide

Research direction

Open review/checklist.md and compare its structure with the deploy guidance in field-notes.md. Inspect how the /review and /ship skills are organized, then add a focused section covering the five listed silent deploy failure cases and clarify its pre- or post-deploy scope. Done means the checklist follows repository conventions and renders correctly.

Written by the indexing model from the issue text.

Assessment

Domain
devops, documentation
Issue type
Documentation
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
65/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.