Menu
Post-deploy verification loop — poll a rollout until every instance is on the new version and healthy, then run one smoke check against a real user path. Meant to be handed to a time-based loop with a timeout, not polled by hand.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Recurring maintenance pass over automated dependency-bump PRs (Dependabot, Renovate, or similar) — rebase, relock, verify, and report what's ready to merge. Meant to be handed to a time-based loop or schedule, not run once.
Reference for designing agent loops — cycles of work that repeat until a stop condition is met. Covers the four loop shapes, writing completion criteria, carrying state and isolating work across cycles, and what running unattended still leaves on you.
Recurring check-in on a stack of dependent PRs — rebase children onto updated parents, surface CI state, flag anything waiting on a human-only gate, and report only what changed since last time. Meant to be handed to a time-based loop, not run once.
Pick the skill or flow that fits your situation. A router over the skills in this repo.
Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes — Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match what the originating issue/PRD asked for?). Runs both reviews in parallel sub-agents and reports them side by side. Use when the user wants to review a branch, a PR, work-in-progress changes, or asks to "review since X".
Shared vocabulary and principles for designing deep modules. Use when the user wants to design or refine a module's interface, hunt for deepening opportunities, decide where a seam belongs, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.
| name | deploy-verify-loop |
| description | Post-deploy verification loop — poll a rollout until every instance is on the new version and healthy, then run one smoke check against a real user path. Meant to be handed to a time-based loop with a timeout, not polled by hand. |
| disable-model-invocation | true |
A deploy that reports success isn't the same as a deploy that's actually healthy — a rollout can finish while a pod crash-loops, a health check flakes, or the new version 400s on a path the deploy step never exercised. This loop watches the window right after a deploy closes and turns "did it work?" into a checked answer instead of a hope.
Before polling, confirm the concrete signal for this deploy target — it varies by platform:
Ready, no CrashLoopBackOff.Done when you can name the exact check that proves the rollout finished — not "give it a few minutes."
Check the Step 1 signal on an interval matched to how fast this deploy actually rolls (a Kubernetes rollout finishes in seconds to minutes; a CDN/edge propagation can take longer — don't poll a slow system every few seconds). Set a hard timeout. If the timeout is reached before the signal clears, stop and report exactly what's still not ready — don't keep polling past the bound hoping it resolves.
A rollout can report "ready" while the new version is still wrong: pull recent logs for the just-deployed instances and scan for a crash loop, an elevated error rate, or a startup panic that a readiness probe doesn't catch. If anything here looks wrong, stop and report it — don't proceed to the smoke check against something already unhealthy.
Exercise a real, critical-path action end to end (not just a /health endpoint returning 200) — whatever this system's core round-trip is: a request that touches the code you just shipped, not a static page. One meaningful check beats a broad shallow sweep here; the goal is "does the thing users actually do still work," not full regression coverage.
Done when the smoke check has run against the newly-deployed instance and you have a pass/fail result, not an assumption.
Do not roll back, restart, or otherwise change the running system as part of this loop — verification surfaces the problem; fixing it is a separate, deliberate decision.