Frontend Secret Exposure Review

    1

    Review whether server-only values reach browser bundles, source maps or web manifests. Get reproducible local evidence and a regression check—without testing real credentials or touching a live site.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Frontend Secret Exposure Review

    Example session with this skill installed

    Synthetic local example: an approved public-output folder contains app.js with the declared public marker and manifest.webmanifest with the declared server-only marker. Use the synthetic contract, run the unchanged checker into a new evidence directory and report artifact evidence. Do not build, deploy, contact a site or validate credentials.

    • Read your context and instructions
    • Compiled the frontend secret exposure

    Recorded synthetic control: exposure-observed (exit 1). One server-only canary finding in manifest.webmanifest at line 1; one public-control observation in app.js; skipped files:
    0. JSON/Markdown reports record artifact/input/runner hashes and match locations without raw canary values. deploymentVerified: false; credentialValidityVerified: false. This is an inert test, not an actual credential leak.

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

    About this skill

    Know what your frontend is shipping—not just whether it builds.

    A server-side value can reach browser output through a public environment prefix, a build configuration change or a source map. A suspicious variable name alone does not prove exposure. This skill gives your coding agent a focused workflow for tracing the boundary and backing its conclusions with local evidence.

    Built for solo developers, small product teams and reviewers checking frontend releases or changes to environment configuration.

    Turn a vague concern into a reproducible check

    • Trace the path: connect a value's intended visibility to source/config and the supplied client artifacts.
    • Check more than the main bundle: inspect supported JavaScript, HTML, CSS, JSON, source maps, SVG and .webmanifest text.
    • Avoid the wrong alarm: distinguish declared public configuration from server-only values; keep unverified classifications explicit.
    • Leave useful evidence: get JSON/Markdown reports with artifact hashes, match locations, public-control observations and coverage limits.
    • Reproduce safely: use inert synthetic canaries in an approved temporary copy instead of putting genuine credentials into a test.

    What you get for $5

    One focused agent skill, a standalone Node.js checker, a synthetic example, an explicit input contract and framework/provider reference notes. The checker uses Node built-ins only: no package installation, account, paid API or live URL is required to inspect approved local build output.

    The agent workflow asks for evidence, a scoped impact explanation, a proposed fix and a regression check—not an unsupported declaration that your application is secure. Source-only risks stay separate from confirmed local artifact observations.

    A concrete example

    Your browser bundle looks clean, but a synthetic server-only marker is present in manifest.webmanifest. The checker identifies the manifest and match location while recording the harmless public marker as a control—not a secret finding. You have a specific artifact to investigate and a check to repeat after a fix.

    Tested on real builds

    Version 0.1.1 has 43 local checks: 29 checker/CLI tests and 14 evidence/infrastructure checks. Six actual Vite 8 builds exercised controlled exposure and no-exposure paths. A separate private React/Vite 7 application passed native TypeScript/build checks in five scratch-copy cases, including manifest-only controls; original project fingerprints stayed unchanged. Test values and injected regressions were synthetic, not customer incidents.

    Scope, clearly stated

    This is an agent review workflow plus an exact synthetic-canary checker—not a universal secret scanner, live-site pentest or credential verifier. It does not discover every unknown secret. No-match covers only the supplied artifacts and supported representations; it is not a security certificate. Application builds require their own dependencies and authorization. Independent agent comparisons, customer benefit and Agensi delivery have not been tested.

    Give your next frontend review a concrete question, local evidence and a repeatable regression check.

    Exact synthetic-canary search, not discovery of all genuine secrets or a security certificate. Supplied visibility and public-output boundaries are not independently verified. Unsupported/binary files, arbitrary transformations, runtime responses and deployed delivery are uncovered. Web manifests are searched as UTF-8 text, not validated as JSON/PWA schema. No live credential, authorization, backend or provider testing. Independent agent benefit, demand and actual Agensi delivery remain untested.

    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 20 days ago

    Requires

    Agensi
    Terminal access
    Node.js 22+
    Local build output
    Synthetic-canary contract
    Node.js

    Permissions

    Read Files
    Write Files

    Allowed Hosts

    mcp.agensi.io

    File Scopes

    Approved
    source/config
    synthetic
    contract
    and
    local
    public-output
    directory:
    read
    only.
    New
    approved
    evidence
    directory
    outside
    the
    inspected
    output:
    create
    report.json
    report.md.
    temporary
    execute
    unchanged
    checker
    as
    .mjs
    remove
    skill
    material
    afterwards.
    Application-copy
    instrumentation
    or
    building:
    only
    with
    separate
    authorization.

    Agensi supplies skill instructions/files; model-provider data handling applies separately. The checker has no network calls, package installation, live-site connection, credential verification or deployment. It does not fetch manifest icons/URLs. Keep source and reports local unless separately approved. Do not read unrelated .env/credentials, modify the working project, test discovered keys, rotate credentials or publish findings automatically.

    Listed20 days ago

    What's inside

    Frequently Asked Questions

    A scoped source-to-artifact workflow, explicit visibility/control rules, a deterministic exact-canary checker, reproducible evidence fields and documented boundaries. It structures the review rather than relying only on suspicious names. No measured advantage over an unassisted agent is claimed.

    The checker reads approved inputs and writes new reports. It never builds, connects, installs, rewrites inputs, validates credentials or deploys. Scratch-copy instrumentation, application builds and fixes require separate authorization.

    No. Agent instructions guide a review, while the helper searches exact synthetic canaries in literal, JSON-escaped, URI-encoded or base64 form. It does not replace generic secret detection or prove all delivery paths were checked.

    No. Intended visibility and provider privileges matter. Declared public markers are controls, not server-secret findings; unknown classification blocks a conclusion. Classifications are supplied, not independently verified by the checker.

    43 local checks, six actual Vite 8.3.5 synthetic-app build controls, and five private React 19/Vite 7.3.6/TypeScript 5.9.3 scratch-copy build cases. Manifest-only public/server controls were separated; original project fingerprints stayed unchanged. These are author-operated synthetic probes, not independent agent/customer or deployment tests.

    No declared server canary was observed in the supplied supported text/representations. Verify the correct build and positive public control. Unsupported files, arbitrary transforms, runtime responses and actual deployment remain outside that result.

    New in Security