More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Code Auditor: Codebase Health Check

    1

    Audit a codebase in three passes: map entry points and data flow, hunt the risks most likely to cause incidents, then deliver a severity-ranked verdict and practical fix sequence.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Code Auditor: Codebase Health Check

    Code Auditor: Codebase Health Check

    Example session with this skill installed

    Audit an inherited subscription billing app before a release decision. Scope: the web API and worker at one commit, two days, source and CI reports available, no penetration test. Known facts: an admin export endpoint was added recently; order creation calls inventory and payment services; two pricing modules exist; the suite is green but several integration tests are skipped.

    • Read your context and instructions
    • Compiled the code auditor

    Verdict: do not release until the export authorization path is verified and fixed if the sampled code matches production configuration. The audit map found five entry points, two privileged roles, one payment path, and one customer export. The dangerous path was traced from request handling through authorization, query construction, file generation, and logging.

    Critical, confirmed — Admin export branch returns before the administrator guard when a CSV format is selected (src/routes/admin.py:88). Any authenticated account that can reach the branch can request the customer export. Immediate mitigation: move the role check ahead of format branching and add a regression test for both formats. Estimated fix: about one hour, plus review.

    High, probable — Order creation decrements inventory and captures payment in separate operations without a transaction or idempotency key (src/orders/service.py:140). A worker crash can leave a paid order without reserved stock. Runtime retry configuration was not in the supplied evidence, so confidence remains probable rather than confirmed. Fix path: make the local state transition transactional and pass a stable idempotency key to payment capture.

    Medium, confirmed — Pricing logic is duplicated (src/orders/pricing.py and src/reports/pricing.py) and the copies already differ on tax rounding. Because one copy feeds invoices, this is scheduled before the release after next, with a shared-module extraction estimated at one day.

    Test integrity: the green suite does not clear these findings. Seven integration tests are skipped, and no test covers export access by a non-admin. Cleared areas: password reset tokens expire correctly in the sampled implementation, and dependency lockfiles are present. Not reached: infrastructure configuration and load behavior. Fix sequence: export guard today, transaction design this week, pricing consolidation next.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    A codebase can pass its tests and still be one export endpoint away from a bad week. This skill runs a bounded audit that starts with evidence, not opinions. First it maps entry points, privileged roles, data stores, dependencies, and change hotspots. Then it hunts in risk order: trust boundaries, authorization, state and concurrency, supply chain, complexity on important paths, and whether the tests actually prove anything. Scanner output is treated as leads, not findings, until reachability and consequences are checked. Every published finding names the location, evidence, realistic impact, confidence, and estimated fix effort. The method includes an intake template, an end-to-end trace for the highest-consequence workflow, a coverage checklist, severity and confidence rubrics, modes for pre-merge and security-only work, and a close-out rule that distinguishes inspected, sampled, and unreached areas. It is built for inherited projects, release decisions, acquisition diligence, and teams that need a clear answer to a practical question: what should we fix first, and what evidence supports that order?

    • Findings are bounded by the snapshot and evidence supplied; omitted runtime configuration can lower confidence or hide defects.
    • It is not a penetration test and does not prove that a system is secure.
    • It does not execute the codebase unless your agent environment separately has that capability and you authorize it.
    • Effort estimates are planning aids based on the visible change surface, not delivery commitments.

    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

    Scanners generate leads. This skill maps the system, verifies reachability and consequences in context, separates confirmed findings from suspicions, and sequences fixes by risk reduced per effort.

    The method is sampling-based, so scope matters more than raw size. Define the decision, snapshot, highest-consequence paths, and time box first; the report must state what was inspected, sampled, and not reached.

    Only if the evidence shows bounded repairs cannot address the risk. The default output is a staged remediation plan with interim mitigations, not a rewrite recommendation based on style or age.

    New in Code Quality & Review