- Home
- Skills
- Testing & Debugging
- deepbug hunter
Works with the AI Agents you already use
deepbug hunter
Systematic bug hunting that proves findings with data flow traces and reproducing unit tests.
$10
deepbug hunter
Example session with this skill installed
Audit the /services/payment-processor module in this repo for logic bugs. I have authorization to test this code.
- Read your context and instructions
- Compiled the deepbug hunter
I have audited the /services/payment-processor module. After mapping entry points and tracing data flows, I identified one confirmed logic flaw regarding floating-point precision during currency conversion.
Confirmed Finding: Precision Loss in Multi-Currency Settlement
| ID | Class | Risk | Path | Status |
|---|---|---|---|---|
| DBH-001 | Arithmetic/Precision | High | POST /settle -> Converter.calculate() -> DB.update() | Confirmed |
Trace Analysis
- Entry:
SettlementController.js:42receivesamountas a float from the request body. - Sink:
CurrencyService.js:108performsMath.round(amount * rate)before passing to the ledger.
Flaw: The rounding occurs before the final summation in the batch processor, leading to a discrepancy of up to 0.04 units per 100 transactions in high-frequency environments.
Reproduction Test
const { calculateSettlement } = require('./services/payment-processor');
const assert = require('assert');
describe('DeepBug Hunter Reproduction: DBH-001', () => {
it('should fail when cumulative rounding errors exceed 0.01', () => {
const rate = 1.255;
const inputs = [10.11, 10.11, 10.11]; // Total expected: 38.07435 -> 38.07
const results = inputs.map(val => Math.round(val * rate * 100) / 100);
const sum = results.reduce((a, b) => a + b, 0);
// The current implementation rounds per-item instead of at the total.
// sum is 38.10, expected is 38.07.
assert.strictEqual(sum, 38.07, 'Cumulative precision error detected in settlement');
});
});
Next steps
- Refactor
CurrencyService.jsto useBigIntor a decimal library for all intermediate calculations. - Move the rounding operation to the final step of the batch settlement process.
- Run
scripts/dedupe_findings.pyafter applying the fix to clear the report.
Connects securely to your tools. The creator never sees your data.
About this skill
Auditing a codebase for subtle logic flaws often results in superficial suggestions rather than actionable fixes. DeepBug Hunter stops the guesswork by enforcing a rigorous verification workflow for every potential vulnerability. It eliminates false positives by requiring a traced data flow and a functional reproduction test for every reported issue.
What it does
- Surface mapping identifies entry points, privileged operations, and risky sinks within the repository.
- Risk prioritization focuses analysis on critical modules like auth, financial logic, and state transitions.
- Hypothesis generation uses structured bug classes to identify specific failure modes in code logic.
- Verification tracing tracks data from input to failure point to confirm if guards are bypassed.
- Minimal reproduction generates failing unit tests to prove the existence of every confirmed bug.
How it works
- Trigger the skill by providing a path to a codebase you are authorized to test.
- The agent maps the surface area and identifies high-risk sinks and entry points.
- The agent formulates hypotheses based on language-specific bug classes and traces data flows.
- You receive a report containing confirmed findings, traced paths, and reproducing unit tests.
Frameworks & tools
This skill works with any codebase where local scripts for mapping and static analysis can be executed. It includes specialized modules for python mapping and uses structured JSON for finding deduplication.
Why this beats prompting it yourself
Standard prompts often hallucinate bugs or miss deep concurrency and state issues. This skill enforces a "no proof, no report" rule, ensuring you only spend time on verified, reproducible flaws rather than generic "best practice" advice.
Use cases
- Audit financial logic for precision errors or race conditions in transaction flows.
- Identify auth bypasses by tracing data from external inputs to privileged sinks.
- Find edge-case state transition bugs in complex backend services.
- Clean up legacy repositories by identifying dead code paths and logic gaps.
Known limitations
Requires authorization to run analysis scripts on the repo. Does not produce weaponized exploits or perform live system attacks.
Requires execution of local Python and Bash mapping scripts. Will not analyze code without explicit authorization. Findings are limited to verifiable logic and data flow flaws.
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 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 1 month ago