Skip to content

Kartik-Doye/Ravengard-AI-Recruiter / .agents/skills/runway

runway

Verified

Deploys software through the target platform's real release model, sequencing configuration, secrets, data, and code safely, then verifying the immutable release and public behavior. Use when the user asks to deploy, release, publish, roll back, or prove a version is live.

FileKindSize
SKILL.mdskill md9.0 KB
references/platforms.mdmarkdown3.6 KB
references/release-protocol.mdmarkdown1.3 KB
references/research.mdmarkdown823 B

references/platforms.md

View on GitHub
# Platform playbook: one methodology, per-platform specifics

The workflow in SKILL.md is identical everywhere; only the commands and
config differ. "What good looks like" per platform: extend this file
as real deployments accumulate (never fabricate from memory; scout
verifies current CLI behavior).

## Vercel (first-class)

- Signals: `vercel.json`, `.vercel/`, Next/Nuxt/etc. framework presets.
- Auth: `vercel whoami` · env: `vercel env ls <env>` (names only) ·
  deploy: `vercel deploy` (preview), `vercel deploy --prod`.
- Previews are the superpower: every branch gets a URL: verify there
  first, always.
- Post-deploy: `vercel ls` confirms the deployment; alias/domains via
  `vercel domains`/`vercel alias`; rollback: `vercel rollback`.
- Migrations: none managed by Vercel: external DB (Neon/Supabase/…)
  applies them as a pre-deploy step with the same safety rules.

## Netlify / Cloudflare Pages (similar shape)

- Signals: `netlify.toml`, `wrangler.toml`/`wrangler.jsonc`.
- Auth + deploy via their CLIs (`netlify deploy`, `wrangler pages
  deploy`); preview deployments per branch; env via dashboard/CLI
  (`netlify env:*`).
- Rollback: previous deployment promote (`netlify restore`,
  wrangler rollback): verify the exact command with scout when first
  used on a project.

## Containers (Docker → anywhere)

- Signals: `Dockerfile`, `compose.yaml`, `.dockerignore`.
- Local build first: `docker build -t app:sha .`: the sha tags the
  exact artifact; the same image that was built is the image that ships.
- Registry push → platform pull (Fly/Railway/Render/ECS/k8s).
- Health check: the container's health endpoint (compose
  `healthcheck`, platform equivalents); rollout status watched, not
  assumed.
- Rollback: redeploy the previous image tag: which is why tags are
  immutable shas, never `latest`.

## VPS / systemd (the boring classic)

- Build artifact or container shipped over SSH; systemd unit with
  `Restart=on-failure`; Nginx/Caddy in front for TLS (Caddy automates
  certs).
- Env via a root-owned env file, never in the unit file or the repo.
- Rollback: previous artifact directory + symlink flip; DB backup
  before migrations (`pg_dump`/`mysqldump` verified restorable).
- Smoke: `curl` the health endpoint through the public domain, verify
  TLS, watch `journalctl` for startup errors.

## Static sites (GitHub Pages / object storage)

- Build output deployed: verify the *generated* files (the HTML
  actually contains the meta tags), not just "upload finished".
- CDN cache: state the invalidation status; stale-cache false alarms
  are the #1 static-deploy confusion.
- Rollback: previous artifact directory/versioned bucket prefix.

## Kubernetes (when it's real)

- Signals: manifests/Helm/kustomize under `deploy/` or `k8s/`.
- `kubectl rollout status deployment/app` is the deploy truth; readiness
  probes gate traffic; `kubectl rollout undo` is the rollback.
- Migrations: init-container/job pattern, never in the app start-up
  path of every replica.

## Mobile (EAS / store pipelines)

- Build via `eas build` (or native pipelines); stores impose review
  timelines that are outside anyone's control: report them as external
  state, never as failures.
- Rollback = phased release pause/halt or store-managed version revert.
- Smoke = the built binary on a device/simulator for the release
  critical paths.

## Universal post-deploy evidence (every platform)

1. version fingerprint proving the new build is live;
2. health endpoint status + timing;
3. 2–3 smoke paths with results;
4. first-minutes log scan (clean or findings);
5. domain/HTTPS status;
6. the rollback command, stated with its preconditions.
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall