More screenshots
Works with the AI Agents you already use
Ship-Safe — Production-Readiness Audit for AI-Built Apps
Run 151 production-readiness checks across 11 dimensions. Get a 0–100 scorecard, file:line evidence, prioritized fixes, and a SHIP / FIX FIRST / DO NOT SHIP verdict before deploying your AI-built app.
$9
Ship-Safe — Production-Readiness Audit for AI-Built Apps
Example session with this skill installed
Audit sample-app for production readiness using the full 151-check checklist.
- Read your context and instructions
- Compiled the ship-safe
Production-Readiness Report
Repo: sample-app · Stack: Node.js/Express, raw pg client, Redis dep (unused), Docker · Date: 2026-11-01
Verdict: DO NOT SHIP
Overall score: 40 / 100 (capped — critical findings present; raw mean 73)
One-sentence summary: a working Express CRUD app with SQL injection on multiple routes, committed and hardcoded secrets, plaintext passwords, an unsigned webhook, and no observability, deploy, or recovery story.
Scorecard
| Category | Score | Findings | Notes |
|---|---|---|---|
| Resilience & Timeouts | 74 | 6 | no timeouts on outbound calls, unbounded in-memory queue |
| Data Layer | 65 | 11 | N+1, no pool, no migrations, unbounded selects, check-then-insert |
| Security | 0 | 19 | SQLi, secrets in repo/code/Dockerfile/logs, no auth, SSRF, XSS |
| Observability | 72 | 10 | console.log only, no health endpoint, no metrics |
| Deployment | 70 | 9 | floating deps, no lockfile/CI, secret in image, no rollback |
| Scaling & Performance | 85 | 7 | in-process queue/state, sequential awaits, no cache |
| Concurrency | 73 | 8 | check-then-act races, no distributed-lock story for cron |
| Async & Messaging | 53 | 11 | unsigned webhook, ack-before-process, no DLQ, in-process cron |
| Reliability & DR | 69 | 7 | no backups, no failover, SPOFs everywhere |
| Infrastructure | 75 | 7 | root container, latest tags, secret in ENV, no IaC |
| Deployment & Ops Context | 67 | 13 | no LB/proxy/runbook/on-call/postmortem evidence |
Checks skipped as N/A: R3, R4, D11, D12, S12, S17, O8, X3, X4, X5, X7, X10, X11, C4, C5, A4, A13, A14, A15, V5, V8, I5, I7, I9, I10, I12, I13, I15, T4, T9, T14
Checks UNVERIFIABLE statically: V2 (prove restores work — run a drill), T8 (verify replica config in cloud console), T18 (org-level on-call), S22 (edge config outside repo)
Findings
| Sev | Check | Finding | Evidence | Fix |
|---|---|---|---|---|
| critical | S1 | .env committed with credentials | sample-app/.env | Remove from history, rotate values, gitignore — security-infrastructure |
| critical | S3 | SQL injection via template literals on 6+ routes | server.js:28, server.js:45, server.js:70, db.js:18, db.js:26, server.js:76 | Parameterize all queries — application-security |
| critical | S7 | Mutating routes have no auth at all | server.js:60 /admin/reset, server.js:65 /charge, server.js:70 /webhook | Auth middleware + authz — application-security |
| critical | S9 | Plaintext password comparison | server.js:29 | bcrypt/argon2 + never store plaintext — security-infrastructure |
| critical | S20 | Passwords and Stripe key written to logs | server.js:27, server.js:82 | Remove; scrub log sinks — observability |
| critical | A10 | Webhook processes body with no signature check | server.js:70-73 | Verify HMAC + timestamp, dedupe by event ID — realtime-apis |
| critical | I8 | Secret baked into image via ENV | Dockerfile:6 | Runtime secret injection — containers-and-iac |
| critical | R1 | Outbound calls have no timeout | server.js:50, server.js:66 | axios.create({ timeout }) + AbortSignal — resilience-patterns |
| critical | V1 | No backup mechanism for the datastore | searched — absent | Automated backups + restore drills — disaster-recovery |
| high | S2 | Secrets hardcoded in source | server.js:16-17 | Env/manager injection — security-infrastructure |
| high | S4 | ORDER BY built from request input | server.js:77 | Allowlist sort columns — application-security |
| high | S5 | Reflected XSS — req.query.name into HTML | server.js:76 | Encode output / CSP — application-security |
| high | S8 | No ownership check on /orders/:id | server.js:44 | Authorize by user ID — application-security |
| high | S10 | JWT secret hardcoded + admin:true for all users | server.js:31 | Env secret, least-privilege claims, full validation — security-infrastructure |
| high | S11 | Access-Control-Allow-Origin: * + credentials | server.js:21-22 | Explicit origin allowlist — application-security |
| high | S14 | Server fetches arbitrary user URL (SSRF) | server.js:49-50 | Scheme/host allowlist, block private IPs — application-security |
| high | S18 | No rate limiting on /login | searched — absent | Limiter on auth + public mutations — traffic-and-edge |
| high | S19 | req.body used unvalidated in queries/logic | server.js throughout | zod/joi schema validation — application-security |
| high | D1 | N+1: per-user orders query in a loop | server.js:39-41 | Eager load/join — database-performance |
| high | D4 | Unbounded SELECT * FROM users endpoint | server.js:38 | Limit/paginate — database-performance |
| high | D6 | Single ad-hoc client, no pool | db.js:7-12 | pg.Pool with sizing — database-performance |
| high | D8 | Read-modify-write with no transaction | db.js:26-30 | Atomic UPDATE ... WHERE stock >= $1 or tx — concurrency-and-memory |
| high | D10 | No migrations at all | searched — absent | Versioned migration tool — versioning-and-migrations |
| high | O7 | No health endpoint | searched — absent | /healthz + /ready — deployment-and-release |
| high | O12 | Request data (incl. passwords) logged | server.js:27 | Structured logs, PII-safe fields — observability |
| high | P1 | Floating deps + no lockfile | package.json (*, latest, >=) | Pin + lockfile — versioning-and-migrations |
| high | P3 | No rollback path | searched — absent | Artifact retention + revert plan — deployment-and-release |
| high | P8 | No CI — nothing gates deploys | no CI config found | CI with test/lint gate — containers-and-iac |
| high | P10 | Secret baked into image (ENV STRIPE_KEY) | Dockerfile:6 | Runtime injection — containers-and-iac |
| high | X1 | Queue/jobs live in process memory | worker.js:3-8 | Durable queue — messaging-and-async |
| high | C1/C2 | Check-then-act races (createUser, decrementStock) | db.js:16-30 | Unique constraint + atomic statements — concurrency-and-memory |
| high | A1 | Consumers process with no dedupe | worker.js | Idempotency by message ID — messaging-and-async |
| high | A2 | ack() before processing | worker.js:12-13 | Ack after success — messaging-and-async |
| high | A3 | No DLQ for poison messages | worker.js | DLQ + redrive — messaging-and-async |
| high | A8 | In-process setInterval cron — double-runs per replica, dies on restart | worker.js:29 | External scheduler or leader lock — messaging-and-async |
| high | A11 | Webhook dedupe missing | server.js:70 | Event-ID idempotency — realtime-apis |
| high | I1 | Container runs as root | Dockerfile — no USER | USER node — containers-and-iac |
| high | V4 | Single instance, no failover | searched — absent | Replica/standby — disaster-recovery |
| high | T5 | Stripe charge never recorded locally | server.js:65-68 | Persist + outbox — distributed-systems |
| medium+ | — | 40 further findings: no CSP, no graceful shutdown, unbounded selects, no eager load, CORS redirect-follow on user URLs, no runbook/on-call/postmortem, node:latest, no .dockerignore, no CDN, SELECT * | see checklist | — |
Unverifiable — check manually
| Check | Why static analysis can't decide | How to verify |
|---|---|---|
| V2 | Can't prove backups restore | Restore drill or PITR in cloud console |
| T8 | Replica config may live outside repo | Check DB provider's console for read replicas |
| T18 | On-call is an org decision | Confirm escalation path exists |
| S22 | Edge/WAF layer outside the repo | Confirm CDN/WAF in front of the app |
Top 5 Fixes
- Parameterize every SQL query + remove committed/hardcoded secrets — Security, S — both are live compromise vectors today
Verify webhook signatures + dedupe by event ID — Async, S — forged payment confirmations are free money for an attacker
3.
Auth middleware on all mutating routes + ownership checks — Security, M — currently anyone can reset sessions or read any order
4.
Add /healthz, structured logging, and a CI test gate — Observability/Deployment, M — you currently can't detect or safely roll back an incident
5. Timeouts on outbound calls + bounded queue + ack-after-process — Resilience/Async, S — the worker silently loses jobs and hung deps hang requests forever
What you're doing right
- Async route handlers throughout — no sync blocking on the request path
- Single shared DB client rather than connection-per-request (
db.js) express.json()default body limit in place- Code is small and readable — fixes are all local changes, no architecture surgery needed
Connects securely to your tools. The creator never sees your data.
About this skill
Your app works locally. Is it ready for production?
Ship-Safe gives your AI agent a structured checklist for reviewing an app before deployment. Built for solo developers, vibe coders, and small teams, it turns a broad “review my code” request into an evidence-backed production-readiness report.
What you get
- 151 checks across 11 audit dimensions
- A 0–100 scorecard
- A SHIP / FIX FIRST / DO NOT SHIP verdict
- File:line evidence for findings, or an explicit note when a required artifact is missing
- Prioritized fixes, including the top five actions to take next
- Checks marked PASS, FAIL, N/A, or UNVERIFIABLE, so missing evidence is not mistaken for a pass
What it reviews
Resilience and timeouts, data access, security, observability, deployment, scaling and performance, concurrency, async processing and messaging, reliability and disaster recovery, infrastructure, and operational readiness.
It looks for issues such as missing outbound timeouts, unsafe queries, exposed secrets, unbounded retries, missing health checks, and absent rollback procedures.
How it works
- Ask your agent: “Audit my app for production readiness.”
- The local prescan gathers signals for 88 grep-able checks, including counter-signals that may indicate an existing control.
- Your agent reviews the evidence against the full 151-check checklist.
- You receive a scored report with findings, limitations, and prioritized fixes.
Prescan matches are candidates for review—not confirmed vulnerabilities. The final audit requires agent judgment.
Included in the pack
- The Ship-Safe audit skill
- The 151-check production-readiness checklist
- A structured report template
- The local
prescan.mjsscript - 20 supporting engineering topic skills
Requirements and permissions
Use with a compatible SKILL.md-capable agent that can read your project files. The deterministic prescan requires Node.js 18+; without Node.js, the agent can review the checklist manually.
The audit workflow reads repository files and runs the local prescan. It does not require access to external accounts or services. The prescan makes no network calls.
What it does not do
Ship-Safe does not automatically patch code, deploy your app, or execute the audited application's code. It is not a penetration test, compliance certification, or guarantee of production safety.
The prescan is deterministic; the full audit depends on your agent's judgment and the evidence available. Runtime behavior and external controls may require separate verification.
See the evidence
The public repository includes audit reports for an intentionally broken demo app and a cleaner reference app, showing both DO NOT SHIP and SHIP outcomes.
Use the report to decide what to fix before release—not as a substitute for tests, monitoring, or human review.
Static review cannot verify every runtime behavior or external control. Prescan matches require agent judgment, and detection coverage varies by language and framework. A SHIP verdict is not a security certification or guarantee of production safety.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 15 days ago