- Home
- Skills
- Agents & Orchestration
- ShipGuard — AI Output Quality & Production Readiness Auditor
More screenshots
Works with the AI Agents you already use
ShipGuard — AI Output Quality & Production Readiness Auditor
AI coding agents can write code incredibly fast—but fast code is not necessarily working code .
$9.50
ShipGuard — AI Output Quality & Production Readiness Auditor
Example session with this skill installed
I've updated the authentication flow to include MFA. Is this ready to ship to production? Audit the changes.
- Read your context and instructions
- Compiled the shipguard
- Generated the document
# SHIPGUARD REPORT: MFA Authentication Flow Evidence recovered for PR #442. The system detected a React/Node.js stack with Jest and Playwright. ## Requirement Traceability Matrix | Requirement | Implementation | Evidence | Status | | :--- | :--- | :--- | :--- | | TOTP Secret Generation | `services/auth.ts:L45` | E2 (Unit Test) | PASS | | Recovery Code Backup | `components/MfaSetup.tsx` | E1 (Static) | INSUFFICIENT | | Rate Limiting on MFA | `middleware/rateLimit.ts` | E3 (Runtime Logs) | PASS | | Session Invalidation | `api/mfa/verify.ts:L89` | E0 (Claim) | **FAIL** | ## Critical Findings - **False Completion:** The `MfaSetup.tsx` component contains a `TODO` for generating backup codes. The UI displays the button, but the handler is an empty arrow function. - **Requirement Drift:** Original spec required SMS fallback; implementation only provides TOTP. - **Security Risk:** No evidence found that existing sessions are invalidated upon MFA enablement, allowing session hijacking for already-logged-in users. ## Evidence Ledger - **E3 (Strong):** Observed 429 Too Many Requests response during brute-force simulation on `/api/mfa/verify`. - **E1 (Weak):** Code structure for TOTP generation follows RFC 6238, but no integration test exercises the full flow. ## Production Readiness Score **Readiness: 45/100** | **Verification Coverage: 60%** ## Ship Decision 🔴 **DO NOT SHIP** **Blockers:** Missing backup code implementation and session invalidation logic. **Next steps** 1. Implement the `generateBackupCodes` logic in `services/auth.ts`. 2. Add an integration test in Playwright to verify the full setup-to-login user flow. 3. Fix the session invalidation vulnerability in the MFA verification endpoint.
MfaSetup.tsx
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
🛡️ ShipGuard Pro — The Quality Gate for AI-Built Software
Don’t trust “Done.” Verify it.
AI coding agents can write code incredibly fast—but fast code is not necessarily working code. A green build doesn't prove a feature works. Passing tests don't prove requirements were satisfied. And an AI agent saying “completed” doesn't mean your product is ready to ship.
ShipGuard Pro is an evidence-first AI verification skill designed to challenge those assumptions before they become production problems.
🔍 What ShipGuard Pro Does
ShipGuard acts like a skeptical principal engineer + QA lead + security reviewer + release engineer inside your AI coding workflow.
Instead of asking “Does the code look correct?”, it asks:
- ✅ Was the actual requirement satisfied?
- 🧪 Is there real evidence that the feature works?
- 🔎 Are tests actually testing behavior?
- 🚨 Are there hidden security or reliability risks?
- 🧩 Did implementation drift from the original requirement?
- 🎭 Is the AI claiming something is complete when it isn't?
- 🚀 Is this genuinely ready to ship?
⚙️ Powerful Verification Modes
- 🔬 Full Audit — Comprehensive project health and production-readiness review.
- 🎯 Task Verification — Verify whether a specific feature or requirement is actually complete.
- 🚦 Pre-Ship Gate — Get a clear SHIP, WARNINGS, DO NOT SHIP, or INSUFFICIENT EVIDENCE decision.
- 🧠 Claim Verification — Challenge claims such as “fixed,” “tested,” “secure,” or “production-ready.”
- 🔀 PR / Diff Gate — Analyze changed code, impact radius, regressions, and merge risk.
📊 Evidence-Based, Not Guess-Based
ShipGuard uses a graded E0–E5 evidence model to distinguish assertions from actual verification.
E0 = AI assertion
E1 = Static evidence
E2 = Automated verification
E3 = Runtime evidence
E4 = Full user-flow verification
E5 = Requirement + implementation + runtime + tests + edge cases
Missing evidence is never treated as a pass. ShipGuard clearly identifies what is verified, what remains uncertain, and what cannot responsibly be confirmed.
🛠️ What It Audits
Depending on your project, ShipGuard can evaluate:
🔹 Functional correctness 🔹 Regression risks
🔹 Test effectiveness 🔹 Security & secret exposure
🔹 UX & accessibility 🔹 Performance signals
🔹 Reliability & error handling 🔹 Observability
🔹 Dependencies & environment 🔹 Documentation
🔹 Requirement drift 🔹 False-completion patterns
🎯 Get a Real Ship Decision
Every serious audit ends with a practical decision:
🟢 SHIP — Critical paths have adequate evidence.
🟡 SHIP WITH WARNINGS — Residual risks remain but no critical blockers were found.
🔴 DO NOT SHIP — A blocker, critical risk, or major unverified requirement exists.
⚫ INSUFFICIENT EVIDENCE — There isn't enough access or evidence to make a responsible decision.
🤖 Built for AI Coding Workflows
ShipGuard Pro is a portable SKILL.md-based agent skill that can work with compatible AI coding environments such as Claude Code, Cursor, Codex, Grok, Gemini CLI, and other skill-compatible agents.
🚫 No API keys required.
📦 Portable skill package.
🧾 Machine-readable evidence ledger.
📋 Markdown reports with optional interactive HTML dashboards.
🔐 Security-conscious verification model.
🧠 Bounded project learning and self-healing.
🚀 Stop Shipping AI Confidence. Start Shipping Verified Software.
Whether you're building a SaaS, API, dashboard, automation, internal tool, or AI-generated application, ShipGuard Pro gives your AI workflow a critical final question:
“You say it's done. Now prove it.” 🛡️
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Runtime verification (E3-E4) requires an execution environment. Visual UI audits are limited to static analysis (E1) without browser access.
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 buying a skill.
Reviews
This skill has been used 4 times.
Trust & safety
Security scan
Security scan passed 2 months ago
Permissions
Allowed Hosts
File Scopes