More screenshots
Works with the AI Agents you already use
Security Audit: Codebase Vulnerability Review
Run codebase security audits whose findings survive scrutiny: a six-field finding contract, coverage-led hunting, independent validation, and severity tied to demonstrated impact, not fear.
$5
Security Audit: Codebase Vulnerability Review
Example session with this skill installed
We're a 6-person SaaS team about to onboard our first enterprise customer, whose security review will ask hard questions. Can you audit our API codebase — about 40 endpoints, token auth, multi-tenant data — and give us a report we can act on and show them?
- Read your context and instructions
- Compiled the security audit
Mode decision first: a report artifact the customer will read means full audit mode, all five phases.
Phase 1, recon: entry points mapped (40 endpoints behind token auth at a gateway), trust boundaries drawn (unauthenticated → authenticated user → tenant member → tenant admin), sensitive flows listed (billing export, invite/role management, webhook intake). Coverage plan issued before any hunting: tenant isolation, authorization on object access, webhook signature handling, export paths — plus an explicit exclusion: the payment provider's dashboard settings, not in the repo, recorded as unexamined.
Phases 2–3, hunting and validation: hunters returned 11 structured candidates. Independent validation killed 6 — three were intended admin authority, two rested on proxy behavior not in evidence, one duplicated another's root cause. Five findings shipped, each with all six contract fields.
Lead finding (High): the invoice-download endpoint authenticates the caller but never compares the invoice's tenant to the caller's tenant. Local reproduction with two fixture tenants: tenant A's token downloaded tenant B's invoice PDF. Fix: one comparison at the resource access in the export handler; regression test asserts a cross-tenant request returns 404.
Phase 5, coverage statement: what was examined, what was excluded, and the hypothesis list with blockers — the part your enterprise customer's reviewers will actually trust.
Connects securely to your tools. The creator never sees your data.
About this skill
Most audit reports fail the same way: inflated severities, checklist deviations dressed up as vulnerabilities, and findings the code owner can dismantle in one question. This skill runs audits that hold up. Every candidate must satisfy a six-field finding contract — attacker principal, accepted input, intended control, crossed boundary, affected resource, concrete result — or it is tracked as a hypothesis, not reported as a finding. Severity is calibrated to what was demonstrated: hypotheses get no severity at all, and a decision table adjusts bands by reachability, whether the control was fully defeated, and blast radius. The full-audit workflow runs five phases — reconnaissance and a coverage plan, coverage-led hunting in waves, independent validation of every candidate, reporting with the smallest effective fix at the last trusted decision point, and an honest coverage statement. It includes a finding template, an audit checklist, evidence standards (static analysis for the code path, local reproduction with dummy data, never probing production), and the anti-patterns that sink audit credibility.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
- It audits what is in evidence: deployment topology, proxy behavior, and provider settings outside the repo are recorded as unknowns and hypotheses, not findings.
- Confirmation stops at the minimum effect proving a boundary violation; it does not develop exploits or chain findings into attacks.
- A scoped or time-boxed pass is partial by design — the coverage statement says what was not examined, and the report never implies otherwise.
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
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 24 days ago
Permissions
No special permissions declared or detected