Skip to content

hashgraph-online/awesome-codex-plugins / plugins/alpoxdev/hypercore/skills/hyper-deploy

hyper-deploy

Flagged

[Hyper] Use this skill when the user asks to deploy an app or service, set up a first deployment, change an existing deployment's configuration, diagnose and fix a failed or unhealthy deployment, or prove a repository is deploy-ready with lint, typecheck, and build checks - on Railway, Cloudflare, a Docker image, docker-compose, or a VPS. It detects the target from repository files, applies the matching target guide bundled in this skill, gates production, rollback, and data actions behind explicit approval, and saves a report under `.hyper/deploy/`. Do not use for application feature work, a CI workflow written with no deployment, or a deployment runbook written as a document.

references/targets/railway.md

View on GitHub
# Railway

Target guide for Railway, read after `rules/target-detection.md` routes here. It covers the current authoring file for new work, the deprecated configuration files that still exist in older repositories, non-interactive deploy and verification commands, diagnostics, and what a rollback does not restore.

## Contents

- Canonical Configuration
- Authentication
- Identify And Deploy
- Success Test
- Status And Logs
- Change Settings
- Rollback
- Failure Table
- Approval Required
- Checked
- Sources

## Canonical Configuration

New work uses Infrastructure as Code: one authoring file at `.railway/railway.ts` (generally available), with `.railway/railway.py` and `.railway/railway.go` in beta. Keep exactly one of the three files, and install the matching package next to it (`npm install railway` for TypeScript, `pip install railway-sdk` for Python, `go get github.com/railwayapp/railway-go-sdk@v0.3.0` for Go). The CLI finds the nearest authoring file by walking up from the current directory, so `--file` is only needed when a second file exists.

```ts
import { defineRailway, project, service } from "railway/iac";

export default defineRailway(() => {
  const web = service("web", {
    build: "pnpm build",
    start: "pnpm start",
    healthcheck: "/health",
    healthcheckTimeout: 60,
    env: { NODE_ENV: "production", STRIPE_SECRET_KEY: preserve() },
  });

  return project("my-app", { resources: [web] });
});
```

| Purpose | Authoring form |
|---|---|
| Build and start | `build`, `start` |
| Health gate | `healthcheck` (path), `healthcheckTimeout` (seconds; platform default 300) |
| Between build and deploy | `preDeploy` (one command string) |
| Scale | `replicas: 3`, or a region map such as `{ "us-west2": 2 }` |
| Source | `source: github("owner/repo", { branch: "main" })`, or `image("ghcr.io/acme/api:1.2.3", { autoUpdates: { type: "patch" } })` |
| Variables | `env: { KEY: "value" }`, cross-service `db.env.DATABASE_URL`, shared `ctx.shared.NAME` → `${{shared.NAME}}`, existing value `preserve()` |
| Storage | `volume("data", { region, sizeMB })` plus `volumeMounts: { "/data": data }` |
| Environment-dependent state | `defineRailway((ctx) => …)` with `ctx.environment`, `ctx.isEnvironment(name)` |

Config as Code (`railway.toml` / `railway.json`) is deprecated. Existing files keep working for services that already use them until **2026-12-01** (hard cutoff). New services cannot opt into it. A service cannot be managed by both systems: `railway config plan` stops and names the service that must be migrated first. Plan migrations with `railway config migrate` (dry run by default), write with `--apply`, and remove the old files only after the cutover with `--apply --delete-files`; `railway config migrate cutover` switches a service off Config as Code and `railway config migrate status` shows what remains. Never hand-write a new `railway.toml`.

Builders: Railpack is the default and needs no configuration for Node, Python, Go, PHP, Java, Ruby, Deno, Rust, Elixir, shell scripts, and static HTML. A `Dockerfile` switches the build to the Dockerfile path and must be named exactly `Dockerfile` with a capital D; any other name or location needs `RAILWAY_DOCKERFILE_PATH`. Nixpacks is in maintenance mode upstream and is not offered here as a choice; do not set it. Railpack settings come from `RAILPACK_*` service variables or its own config file.

## Authentication

Two token types exist and only one may be set at a time.

| Variable | Scope | Use |
|---|---|---|
| `RAILWAY_TOKEN` | one project and one environment | deploy, redeploy, logs, variables, `config apply` for that environment |
| `RAILWAY_API_TOKEN` | account and workspace | creating or deleting environments, work across several projects, the MCP server |

Setting both makes the CLI error out. The MCP server rejects project tokens because it needs a user identity for billing and audit, so an MCP-driven run needs `RAILWAY_API_TOKEN`. Never pass a token as a command-line argument; it lands in shell history and the process list.

```sh
[ -n "${RAILWAY_TOKEN:-}" ] && echo project-token-set
[ -n "${RAILWAY_API_TOKEN:-}" ] && echo account-token-set
railway whoami
```

The presence test prints nothing sensitive; `railway whoami` confirms which identity the CLI resolved. Do not run `railway whoami` before the token check when both are cheap, since an unauthenticated call fails instead of telling you which variable is missing.

## Identify And Deploy

Start with the context, not with a deploy:

```sh
railway status --json
```

Then name the project, environment, and service explicitly on every writing command instead of relying on the linked directory. `--project` requires `--environment`; pass `--service` too when the project has more than one.

```sh
railway up --project <PROJECT_ID> --environment <ENV> --service <SERVICE> \
  --detach --json -m "<release summary>"
railway up --project <PROJECT_ID> --environment <ENV> --service <SERVICE> \
  --ci -m "<release summary>"
```

- `--detach` (alias `--no-wait`) returns once the upload is accepted and the build is **queued**. With `--json` it prints `{"deploymentId": …, "logsUrl": …}`; keep the ID.
- `--ci` streams build logs only and exits with the deployment's verdict as its exit code. `--json` implies CI behaviour for an authenticated deploy.
- Outside CI mode, the CLI also returns right after upload when stdout is not a terminal. Omitting `--detach` in an agent run does not mean the command waited; poll for the ID either way.
- A non-interactive run with no linked project fails with `NOT_AUTHENTICATED` or a no-project error instead of creating one. Scripts and CI never create projects implicitly, so the project must exist and be named.

`railway deploy` and `railway down` are not code deploys and are never used for a release: `railway deploy` provisions a template into the project (`--template postgres`), and `railway down` removes the most recent deployment from the service. Code goes out with `railway up` (or the equivalent `railway deployment up`).

## Success Test

Success is this run's `deploymentId` reaching `SUCCESS`, a quiet runtime log window, and a 2xx from the service URL.

```sh
railway deployment list --project <PROJECT_ID> --environment <ENV> --service <SERVICE> \
  --limit 10 --json
railway logs <DEPLOYMENT_ID> --lines 100 --json
curl -sS -o /dev/null -w '%{http_code}\n' https://<domain>/health
```

- Match on the ID returned by `railway up`, never on "the newest deployment". A concurrent newer deploy from another source is not this run's result.
- `--ci` exiting `0` is not sufficient on its own: a build that matches no watch pattern can end as `SKIPPED` with exit `0`, and a detached or non-TTY run can exit `0` after upload alone.
- States run `QUEUED` → `INITIALIZING` → `BUILDING` → `DEPLOYING` (plus `WAITING` behind GitHub checks) → `SUCCESS`. `FAILED`, `CRASHED`, `SKIPPED`, `SLEEPING`, `NEEDS_APPROVAL`, `REMOVING`, and `REMOVED` are not success; report the state as it is and the next action. `NEEDS_APPROVAL` waits for a human and must not be polled as if it were still building.
- The health check is a one-time gate at deploy, not continuous monitoring. A service with an attached volume cannot be zero-downtime even with a health check, because only one deployment may hold the volume.

## Status And Logs

Every log command needs a bound; without `--lines`, `--since`, or `--until` the CLI streams until interrupted.

| Question | Command |
|---|---|
| Current context and resources | `railway status --json`, `railway environment config --json` |
| Deployment history and states | `railway deployment list --limit 10 --json` |
| Build log | `railway logs --build --lines 400 --json` |
| Runtime log for one deployment | `railway logs <DEPLOYMENT_ID> --lines 200 --json` |
| Runtime log of the newest deployment, even a failed one | `railway logs --latest --lines 200 --json` |
| Error-only window | `railway logs <DEPLOYMENT_ID> --lines 200 --filter "@level:error" --json` |
| HTTP failures | `railway logs --http --status ">=400" --lines 100 --json` |
| Dropped network flows | `railway logs --network --status dropped --lines 100 --json` |
| Failed DNS lookups | `railway logs --dns --status failed --since 1h --lines 100 --json` |
| CPU, memory, volume, HTTP metrics | `railway metrics --since 1h --json`, `--memory`, `--http`, `--raw` |
| Domain and certificate state | `railway domain list --json`, `railway domain status <domain> --json` |
| Variables | `railway variable list --json` (prints values; do not paste the output) |
| Cron runs | `railway cron list` |

`railway logs` accepts HTTP, network, and DNS filters (`@httpStatus:>=500`, `@path:/api`, `@peer_kind:internet`, `@rcode:NXDOMAIN`) and one log type per call. The deploy log defaults to the most recent successful deployment, so pass the ID when diagnosing a failure.

## Change Settings

Variables are staged: setting one records a pending change that reaches the service on the next deploy. `railway variable set` triggers that deploy unless `--skip-deploys` is set.

```sh
railway variable set DATABASE_URL --service <SERVICE> --environment <ENV> --stdin
railway variable set LOG_LEVEL=debug --service <SERVICE> --environment <ENV> --skip-deploys
railway variable delete OLD_KEY --service <SERVICE> --environment <ENV>
```

- `--stdin` reads the value from stdin, so the secret never appears in the argument list. `KEY=VALUE` on the command line is for non-secrets only.
- `--skip-deploys` records the change without deploying; the value is not live until a deploy runs.
- `variable list --json` and `-k/--kv` print raw values. Read them only when needed and never copy them into the report.
- Sealed variables are writable but their value is never readable: they cannot be un-sealed, are not returned by `railway variables` or `railway run`, are not copied into pull-request environments or when duplicating an environment or service, are not shown in environment diffs, and do not sync with external integrations. A duplicate environment therefore arrives without them, and a run that assumes otherwise fails at startup.

Infrastructure changes go through plan and apply, never a hand edit of live settings:

```sh
railway config plan --environment <ENV> --detailed-exit-code
railway config plan --out railway-plan.json
railway config apply --plan railway-plan.json --yes --confirm-destructive
```

- `plan` only reads; variable values appear as `«hidden»` unless `--show-values` is passed.
- `--detailed-exit-code` makes `plan` exit `2` when changes are pending and `0` when none are, which is the CI drift gate.
- A non-interactive apply of a destructive change needs both `--yes` and `--confirm-destructive`. `apply` always runs a fresh plan and rejects a plan whose environment etag or `.railway/` tree no longer matches, so re-plan after any concurrent change.
- `railway config pull --include-variables` decrypts non-sealed values, including unsealed secrets, into `.railway/railway.ts`. Never run it in an agent flow; keep imported values as `preserve()`.

## Rollback

Rolling back to an arbitrary earlier deployment is a dashboard action; the CLI has no rollback subcommand.

- A rollback restores the previous Docker image **and** the custom variables from that deployment, so a variable changed since then is reverted too.
- It only works inside the plan's image retention window: Free and Trial 24 hours, Hobby 72 hours, Pro 120 hours, Enterprise 360 hours. Past that the option disappears and only a redeploy remains.
- It restores code and service variables only. Databases, volumes and their data, buckets, and any resource bound to the service are not restored, and a schema migration that already ran stays run.
- Before proposing one, state the retention window, that variables revert, and that data does not.

Alternatives, in order of preference when the cause is not the image:

| Command | Rebuilds | Use when |
|---|---|---|
| `railway redeploy --service <SERVICE> --yes` | no, same image | the build is fine and the container needs to restart from a known image |
| `railway redeploy --from-source --yes` | yes, latest commit or image | the failure is stale code |
| `railway redeploy --commit <SHA> --yes` | yes, that commit once | a specific commit must be tried without pinning it |
| `railway restart --service <SERVICE> --yes` | no, same container | the process is wedged at runtime |

`railway down` removes the most recent deployment; it is not a rollback and does not return the service to an earlier image.

## Failure Table

| # | Signature | Cause | Fix | Check | Tier | Source |
|---|---|---|---|---|---|---|
| 1 | `Application failed to respond` (502), health check reporting `service unavailable` or `status 400` | App bound to `localhost` or a hard-coded port; target port differs from the listening port | Bind `0.0.0.0` and read `PORT`; align the domain target port or set a `PORT` variable | `railway logs --http --status ">=400"`, then `curl -I https://<domain>/health` returns 2xx | proceed | <https://docs.railway.com/networking/troubleshooting/application-failed-to-respond> |
| 2 | `No start command could be found` | Railpack could not detect a start command | Set `start` in the authoring file or the start command in service settings (for example `uvicorn main:app --host 0.0.0.0 --port $PORT`) | redeploy reaches `SUCCESS` and the app log shows it listening | proceed | <https://docs.railway.com/deployments/troubleshooting/no-start-command-could-be-found> |
| 3 | Deployment ends `FAILED` around 300 seconds | Health check timed out, or `healthcheck.railway.app` is not allowed by the app's host filter | Raise `healthcheckTimeout` or `RAILWAY_HEALTHCHECK_TIMEOUT_SEC`, remove startup delay, allow the health check host | `SUCCESS` plus a 2xx on the health path | proceed | <https://docs.railway.com/deployments/healthchecks> |
| 4 | `KeyError`, `undefined`, or a connection to an empty host | Variable missing from the environment, or set but never deployed | Set it with `--stdin`, then deploy so the staged change lands | runtime log starts clean, HTTP log shows 200 | proceed | <https://docs.railway.com/variables> |
| 5 | Dependency install fails, wrong package manager, `Cannot find module` | Builder misdetected the project, or the build ran from the wrong directory | Set `build` explicitly, and `RAILPACK_*` variables for the runtime version; move the service root rather than the whole repo | rebuild log shows the intended command, then `SUCCESS` | proceed | <https://docs.railway.com/builds/railpack> |
| 6 | Container exits immediately, `exec format error`, or a restart loop with no app log | Image built for the wrong CPU architecture (arm64 built locally, deployed to amd64) | Rebuild with `docker build --platform linux/amd64`, push, then `railway redeploy --from-source` | `SUCCESS` and an app log line | proceed | <https://docs.railway.com/guides/private-container-registry> |
| 7 | Exit code 137, repeated restarts | Out of memory | Reduce memory use or raise replicas within the plan | `railway metrics --since 1h --memory` stable and no restart for 30 minutes | explicit approval (cost) | <https://docs.railway.com/deployments/troubleshooting/slow-deployments> |
| 8 | `Pre-deploy command timed out after N seconds`, deploy stops | Migration failed or hung; pre-deploy has no volumes and needs its own dependencies | Fix the command, keep it under the 1–3600 second cap, do not touch the volume there | redeploy reaches `SUCCESS`, migration log exits 0 | explicit approval (schema change) | <https://docs.railway.com/deployments/pre-deploy-command> |
| 9 | `ENOTFOUND redis.railway.internal` | Client resolves A records only in an IPv6-only legacy environment, or the service is in another project or environment | Set `family=0` for ioredis and bullmq, confirm same project and environment, use the public URL from outside | `railway logs --dns --status failed` empty and the app connects | proceed | <https://docs.railway.com/databases/troubleshooting/enotfound-redis-railway-internal> |
| 10 | Deployment is `SKIPPED`, "no changed files match watch patterns" | Watch paths excluded the changed files, or a failing GitHub check suite skipped it | Fix watch paths, or `railway redeploy --from-source` | a new deployment reaches `SUCCESS` | proceed | <https://docs.railway.com/deployments/github-autodeploys> |
| 11 | Custom domain returns 404, certificate error | Only the CNAME was added; the TXT verification record is missing, or the certificate needs a retry | Add both records exactly as shown, then `railway domain certificate retry` | `curl -I https://<domain>` returns 200, `railway domain status` shows a valid certificate | proceed | <https://docs.railway.com/networking/domains/working-with-domains> |
| 12 | `EACCES` on writes, or data gone after a deploy | Volume mounted at the wrong path, written by a non-root user, or written during build or pre-deploy | Mount under `/app/...`, set `RAILWAY_RUN_UID=0` for a non-root image, write at runtime only | restart and confirm the file persists | explicit approval (volume change) | <https://docs.railway.com/volumes> |
| 13 | Cron never runs, previous run still active | The process does not exit; schedule is closer than 5 minutes or not in UTC | Make the job terminate and close its connections, use a schedule at least 5 minutes apart in UTC | `railway cron list` shows a completed run at the next slot | proceed | <https://docs.railway.com/cron-jobs> |
| 14 | Exit code `0` but nothing deployed | Upload only, detached or non-TTY return, or a skipped build | Apply the success test above against this run's deployment ID | `deployment list --json` shows that ID as `SUCCESS` | proceed | <https://docs.railway.com/cli/up> |

## Approval Required

| Action | Tier |
|---|---|
| `railway status`, `deployment list`, `logs` with a bound, `metrics`, `variable list`, `domain status`, `cron list`, `railway config plan`, editing the authoring file in the repository | proceed |
| `railway up`, `redeploy`, `restart`, a variable change, `railway config apply` in a non-production environment | proceed |
| The same in production | named environment |
| Production rollback (dashboard), `railway down`, service, environment, domain, volume, or bucket deletion, `railway config apply --confirm-destructive` on a production environment, `railway config partials release` or `transfer`, a pre-deploy or manual database migration against real data, an action that adds cost (replicas, plan, region), creating a new public domain or TCP proxy, `config pull --include-variables` | explicit approval |
| Logging in or creating an account or project, installing a CLI globally, printing a token or a `variable list --json` value, unsealing a sealed variable | never |

Approval covers the one action it names. A rollback approval does not cover the deploy that follows it, and a variable approval does not cover the redeploy.

## Checked

> Volatile. Checked 2026-11-07; recheck the live page when a flag, a limit, or the deprecation date decides the run.

| Fact | Value |
|---|---|
| CLI | railway 5.71.0 installed and run locally on 2026-11-07 |
| Flags | every flag used here was re-read from `railway <command> --help` on that binary, not recalled |
| Authoring file | `.railway/railway.ts` (GA); `.py` and `.go` beta; exactly one file |
| Config as Code cutoff | 2026-12-01, hard; new services cannot opt in |
| Builders | Railpack default, Dockerfile when the file is named exactly `Dockerfile`; Nixpacks not offered |
| Health check | 2xx polling at deploy only, default timeout 300 seconds, host `healthcheck.railway.app` |
| Image retention | Free 24 h, Hobby 72 h, Pro 120 h, Enterprise 360 h |
| Pre-deploy | no volumes, private network available, timeout 1–3600 s, no retry |
| Cron | UTC, minimum 5 minutes apart, process must exit |

## Sources

> Checked 2026-11-07.

| Claim | Source |
|---|---|
| Config as Code is deprecated, works for legacy services until 2026-12-01, and new services cannot opt in | <https://docs.railway.com/config-as-code> |
| IaC replaces Config as Code; one authoring file; a service cannot be managed by both; `config plan` stops on legacy services | <https://docs.railway.com/infrastructure-as-code> |
| TypeScript DSL: services, build, start, healthcheck, preDeploy, replicas, env, preserve, volumes | <https://docs.railway.com/infrastructure-as-code/reference> |
| `config migrate` flags, cutover, undo, status; `plan --out` then `apply --plan`; `--detailed-exit-code`; redacted values; stale-plan rejection | <https://docs.railway.com/cli/config> |
| Railpack builds with zero configuration and covers the listed languages | <https://docs.railway.com/builds/railpack> |
| The Dockerfile must be named `Dockerfile` with a capital D; `RAILWAY_DOCKERFILE_PATH` for other names | <https://docs.railway.com/builds/dockerfiles> |
| Build and start command overrides; exec-form commands need a shell for `$PORT` | <https://docs.railway.com/builds/build-and-start-commands> |
| `RAILWAY_TOKEN` is project-scoped, `RAILWAY_API_TOKEN` account-scoped, and only one may be set | <https://docs.railway.com/cli/login> |
| Project tokens are scoped to one environment and only deployment actions | <https://docs.railway.com/cli/deploying> |
| The MCP server rejects project tokens | <https://docs.railway.com/ai/mcp-server> |
| `up` flags, `--detach` returning at queue time, non-interactive runs never creating projects, exit code meanings | <https://docs.railway.com/cli/up> |
| Early return when stdout is not a terminal and CI mode is off | <https://raw.githubusercontent.com/railwayapp/cli/master/src/commands/up.rs> |
| `deployment list` with `--limit` and `--json`; `deployment up`, `deployment redeploy` | <https://docs.railway.com/cli/deployment> |
| Log types, `--lines` disabling streaming, filter syntax, `--latest` | <https://docs.railway.com/cli/logs> |
| Deployment states and what Active means with and without a health check | <https://docs.railway.com/deployments/reference> |
| Health check polls 2xx, default 300 s timeout, no monitoring after the deploy, volume services take downtime, health check host name | <https://docs.railway.com/deployments/healthchecks> |
| Pre-deploy runs between build and deploy, no volumes, private network available, no retry, 1–3600 s cap | <https://docs.railway.com/deployments/pre-deploy-command> |
| Variables are staged and deployed; sealed variable constraints | <https://docs.railway.com/variables> |
| Staged changes must be reviewed and deployed | <https://docs.railway.com/deployments/staged-changes> |
| Rollback restores the image and custom variables, is limited by plan retention, and is dashboard-only | <https://docs.railway.com/deployments/deployment-actions> |
| Retention windows per plan | <https://docs.railway.com/pricing/plans> |
| Volumes mount at runtime only, not at build or pre-deploy, and `RAILWAY_RUN_UID=0` for non-root images | <https://docs.railway.com/volumes> |
| Custom domain needs CNAME and TXT, a missing TXT returns 404, propagation up to 72 hours | <https://docs.railway.com/networking/domains/working-with-domains> |
| Cron schedules are UTC, at least 5 minutes apart, and the process must exit | <https://docs.railway.com/cron-jobs> |
| Watch paths can skip a deploy, and a failing check suite skips it immediately | <https://docs.railway.com/deployments/github-autodeploys> |
| Exit 137 is an OOM kill | <https://docs.railway.com/deployments/troubleshooting/slow-deployments> |
| arm64 images built on Apple Silicon crash on Railway without `--platform linux/amd64` | <https://docs.railway.com/guides/private-container-registry> |
| ioredis needs `family=0` against the IPv6-only legacy private network | <https://docs.railway.com/databases/troubleshooting/enotfound-redis-railway-internal> |
| `Application failed to respond` means the app is not listening on `0.0.0.0:$PORT` or the target port is wrong | <https://docs.railway.com/networking/troubleshooting/application-failed-to-respond> |
| `No start command could be found` means Railpack found no start command | <https://docs.railway.com/deployments/troubleshooting/no-start-command-could-be-found> |
| Never report a deploy as successful without observing `SUCCESS` for that deployment; `--detach` means queued; exit 0 can be a skip | <https://github.com/railwayapp/railway-skills> |

### Evidence grade

`DIRECT` for the CLI surface: railway 5.71.0 was installed and every command and flag above was read from its `--help` on 2026-11-07, and the non-TTY early return was confirmed in the published CLI source. `SECONDARY` for platform behaviour that only a live project can show, such as retention windows and per-plan limits, which are quoted from Railway's own documentation and were not exercised against a project in this environment. The 2026-12-01 cutoff is a documentation banner, not a signed contract; re-read the page before a migration depends on the date.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 322kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 197kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall