More screenshots
Works with the AI Agents you already use
Check Your App for Leaks Before You Ship
Three checks on your own app before release, in the right order: leaked keys first, then the app's weak spots, then decoys that waste an attacker's time.
$12
Check Your App for Leaks Before You Ship
Example session with this skill installed
Run a pre-ship security pass on the current release candidate. Scan the git history for leaks and audit the new /api/v2/admin endpoints. Generate a findings report with live status.
- Read your context and instructions
- Compiled the check your app
- Generated the document
I have completed the three-layer security pass for the release candidate.
Pre-Ship Security Report
Layer 1: Secret Scan Results
| Finding | Source | Type | Status | Action Taken |
|---|---|---|---|---|
sk_live_... | .env.backup (Commit a1b2c3) | Stripe API Key | LIVE | Revoked and Rotated |
AKIA... | tests/fixtures.json | AWS Access Key | REVOKED | Removed from history |
DB_PASS | config/db.js | Placeholder | FALSE POSITIVE | Noted as example |
Layer 2: Application Audit
- Auth/Sessions: Found
/api/v2/admin/statusaccessible without valid JWT. Action: Added middleware check. - Inputs:
POST /api/v2/admin/usersaccepted 50MB payloads. Action: Implemented 1MB limit. - Endpoints: Leftover
/debug/varsfrom development was active. Action: Removed from production build. - Dependencies:
node-sassversion has a known high-severity CVE. Action: Updated tosass.
Layer 3: Hardening Deliverables
- Fake Credentials: Injected 5 inert API keys into
dist/assets/config.json. - Decoy Interfaces: Added
InternalSchemaManagerandLegacyAuthBridgemodules to the build tree. - Verification: Ran 42 functional tests against the hardened build. Result: 100% Pass (No regression).
Next steps
- Review the new
secrets.json(stored outside the repo) for the rotated Stripe key. - Confirm the
/api/v2/admin/statusfix in the staging environment. - Proceed to deploy the hardened artifact located in
/build/hardened-release.zip.
hardened-release.zip
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
AI-written code ships fast. That is the problem. An agent will happily paste a real key into a test file, leave a debug endpoint open, and commit the whole thing — and the commit history keeps the mistake even after the file is fixed. This pass runs before release, in a fixed order, so the cheap leaks are closed first. Scope, stated in the skill's first section: this is for defending your own app — code you own or are authorized to ship. It does not apply to scanning or probing someone else's system.
Inside the package: the skill file covering all three layers, a secret-scan and hardening checklist you can run as a list, and a filled example pre-ship report. Layer one is secrets, because a live key in the repo is an emergency already in progress: scan the working files and the full git history — every commit, including deleted files — then verify each finding instead of just listing it, classify it as a real credential, a test value, a placeholder, or a public identifier that only looks secret, and check whether real ones are still live. Layer two is the app itself: who can do what without logging in, what a low-privilege account can reach, what happens with oversized or hostile input, which endpoints the shipped build actually serves versus what you meant to serve, known holes in the library versions you pin, and responses or logs that hand back more than the caller needs.
A run goes like this: freeze the release candidate — no new features mid-pass. Layer one scan, verification, and rotation: live means revoke and rotate now, store the replacement outside the repo, and rewrite the history that exposed the old one, because a rotated key left in a public commit is still a leak. Layer two audit, fixes in severity order, each re-tested, and anything you deliberately accept gets recorded with its reason — an accepted risk with a reason is a decision; the same risk undocumented is a surprise for later. Only then, layer three: harden the shipped build with decoys — well-formed but worthless fake keys placed where scrapers and automated tools look first, plausible dead-end interfaces that are not the real implementation, structure that burns an automated analysis budget — while real users get the identical app, proven by running your own full test suite against the hardened build and requiring the same result. Decoys must never be real, and you keep a clean internal build for debugging; the release record notes which artifact is which. The pass ends in a report where every finding has a disposition. "Found 40 things" is not a result. "Found 40, 2 were live, both rotated" is.
Only for use on codebases you own or are authorized to ship. Does not replace server-side security. Hardening must be verified against your own test suite to ensure no behavioral changes.
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 6 days ago