Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Deep Testing

    1

    Rigorous pre-production testing agent that hunts logic bugs and builds hardened test suites.

    Free

    Security scanned
    Deep Testing

    Deep Testing

    Example session with this skill installed

    Test the user authentication logic in `auth.ts`. Find any edge cases and verify it's ready for production.

    • Read your context and instructions
    • Compiled the deep testing

    I have completed a deep audit of the authentication logic in `auth.ts`. I identified a critical flaw in session expiration handling and a medium risk in the password reset token window. ## 🧪 Testing report: auth.ts ### 🚦 Verdict **🟡 Ready with caveats** Logic is sound for standard flows, but session cleanup fails under specific clock skew conditions. | Metric | Value | |---|---| | 🐛 **Bugs found** | 2 (🔴 1 · 🟠 0 · 🟡 1 · ⚪ 0) | | 🔧 **Fixed** | 1 | | ⏳ **Need your decision** | 1 | | 🧪 **Tests added** | 5 (3 unit · 2 property) | | ✅ **Suite** | 12/12 passing → 17/17 passing | ### 🗺️ Coverage map | Area | Tier | Result | |---|---|---| | `login()` | 🔬 Full depth | Clean | | `validateSession()` | 🔬 Full depth | 1 bug found, fixed | | `resetPassword()` | 👁️ Representative | 1 bug found (reported) | ### 🐛 Bugs #### 🔴 #1 Session Valid During Clock Skew · `auth.ts:42` - **Status:** ✅ Fixed - **What:** Sessions remain active if the server clock drifts backwards after the token is issued. - **Example impact:** Token issued at 12:00 (exp 13:00) → Server clock drifts to 11:55 → Session effectively gains 5 minutes. - **Before:** ```typescript const isValid = Date.now() < session.expiresAt; ``` - **After:** ```typescript const isValid = Date.now() < session.expiresAt && Date.now() > session.issuedAt; ``` - **Proof:** `test_session_not_valid_in_past` ❌ before → ✅ after #### 🟡 #2 Token Window Overlap · `auth.ts:88` - **Status:** ⏳ Needs your decision - **Issue:** New reset tokens don't invalidate old ones immediately, allowing multiple active reset links. - **Recommendation:** Invalidate all previous tokens for a user ID upon new request. **Next steps** - Review the recommendation for the password reset token invalidation. - Run `npm test` to verify the 5 new regression tests in your environment. - Merge the fixes in `auth.ts`.

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

    About this skill

    Stop shipping code that only works on the happy path. This skill acts as a rigorous pre-production gate by hunting for bugs that simple unit tests and manual reads miss. It assumes every piece of code has blind spots and uses adversarial review, property-based testing, and fuzzing to find them. ## What it does - **Adversarial Bug Hunting**: Generates hypotheses for failures in edge cases like race conditions, timezone shifts, and malformed inputs. - **Deep Coverage Mapping**: Audits every file in the requested scope using risk-based tiers, ensuring no module is silently skipped. - **Test Suite Hardening**: Writes unit, integration, and property-based tests that target public interfaces rather than internals. - **Automated UI Verification**: Performs end-to-end browser walkthroughs to catch state sync issues, responsive layout breaks, and stale data. - **Regression Proofing**: Delivers targeted fixes for confirmed bugs alongside tests that prove the fix works and stays fixed. ## How it works 1. **Reconnaissance**: Scans the stack to identify test runners, environment configs, and the existing baseline. 2. **Risk Scoring**: Maps the scope and assigns depth tiers based on complexity and impact. 3. **Attack Phase**: Executes a bug hunt using an edge-case checklist and adversarial subagents to break logic. 4. **Verification**: Proves findings with failing tests, applies fixes, and generates a structured testing report. ## Frameworks & tools Works with any language or framework. Uses ecosystem defaults like Playwright or Cypress for E2E, and property-based libraries such as Hypothesis (Python), fast-check (JS), or RapidCheck (C++). ## Why this beats prompting it yourself Standard prompts often produce "happy path" tests that mirror the code's existing logic. This skill uses a tiered risk model and mutation checks to ensure tests actually fail when code is broken, preventing the false security of high coverage numbers. ## Use cases - Hardening a financial calculation engine against rounding errors. - Auditing a multi-tenant auth flow for data leakage. - Verifying a complex UI state machine across mobile and desktop. - Making a legacy module production-ready before a major release. ## Known limitations Does not handle load testing or performance infrastructure. Requires an existing or configurable test harness to execute verification steps.

    Does not handle load or performance testing. Requires a runnable test environment. Cannot perform E2E browser tests if MCP or browser tools are unavailable.

    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

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 day ago

    • Free with an Agensi account

    Requires

    Node.js
    Git

    Permissions

    Terminal / Shell
    Browser
    Read Files
    Listed1 day ago

    What's inside

    Frequently Asked Questions

    Unlike standard prompts that confirm existing logic, this skill uses adversarial subagents and property-based testing to actively try to break your code, finding race conditions and edge cases that simple unit tests miss.

    The skill is language-agnostic and works with any stack, utilizing ecosystem-standard tools like Playwright or Cypress for UI and libraries like Hypothesis or fast-check for logic verification.

    No, this skill is designed to integrate with your existing environment; it scans your local stack to identify your current test runners and configurations before beginning the attack phase.

    You receive the full testing agent skill, including the Risk Scoring engine and the Adversarial Bug Hunting module, along with any logic updates and version improvements released by the developer.

    This skill focuses on logic, UI state, and security hardening; it does not currently handle high-volume load testing, stress testing, or performance infrastructure benchmarking.

    It follows a tiered risk-based approach, auditing every file in your specified scope and assigning 'depth tiers' to ensure complex modules get the most rigorous scrutiny while maintaining efficiency.

    New in Testing & Debugging