More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Security Audit: Codebase Vulnerability Review

    1

    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

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Security Audit: Codebase Vulnerability Review

    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.

    • 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. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 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 24 days ago

    Permissions

    No special permissions declared or detected

    Listed24 days ago
    Updated21 days ago

    What's inside

    Frequently Asked Questions

    Scanner output is a candidate list, not findings. This skill walks every candidate through the six-field contract — who attacks, what control was supposed to stop them, what boundary actually fell — and independently validates each one, so the report contains only claims that survive the code owner's scrutiny.

    Yes, if it presents itself honestly. The workflow audits by data sensitivity first — credentials, money, other tenants' data — and the coverage statement names exactly what was and wasn't examined. A partial audit that says so beats a shallow pass over everything.

    It specifies the smallest effective fix for each finding, located at the exact decision point (file and line) with a regression test to assert — narrow enough for your team to apply and review quickly. Applying changes stays with the code owners.

    New in Security