- Home
- Skills
- Technical Documentation
- Docs Bug Bounty
Works with the AI Agents you already use
Docs Bug Bounty
A technical documentation auditor that finds broken code, API drift, and logical dead ends using a bug-bounty framework.
$24
Docs Bug Bounty
Example session with this skill installed
Run a bug bounty audit on the attached README.md and stripe-style-api.yaml. Focus on the 'Creating a Charge' section. The audience is Node.js developers. Trace code samples by reading.
- Read your context and instructions
- Compiled the docs bug bounty
Audit complete for the 'Creating a Charge' section: 1 Blocker, 2 High.
Docs Bug Bounty Report
VERDICT: BLOCKED
Scope Statement
- IN SCOPE: README.md (Creating a Charge), code samples
- OUT OF SCOPE: Dashboard/UI docs
- VERSION AUDITED: v2024-10-01
- SOURCE OF TRUTH: stripe-style-api.yaml
- EXECUTION: Traced by reading
Leaderboard
| Hunter | Blocker | High | Med | Low | Points |
|---|---|---|---|---|---|
| Example Hunter | 1 | 1 | 0 | 0 | 150 |
| Drift Hunter | 0 | 1 | 0 | 0 | 50 |
| Total risk score | 200 |
Findings
E-01 - Client never initialized (Blocker)
- LOCATION: README.md, Creating a Charge, code block 1
- REPRODUCTION: Copy the snippet and run it - throws
ReferenceError: client is not defined(traced) - EVIDENCE: The snippet calls
client.charges.create(...)but no setup line appears anywhere in the section - FIX: Add
const client = new Payments(process.env.PAYMENTS_TEST_KEY);above the call
E-02 - Top-level await outside an async context (High)
- LOCATION: README.md, Creating a Charge, code block 1
- REPRODUCTION: Running the file as CommonJS fails with
SyntaxError: await is only valid in async functions(traced) - FIX: Wrap the call in
async function main() { ... } main();or note that the sample requires an ES module
D-01 - Parameter renamed in schema (High)
- LOCATION: README.md, Creating a Charge, parameter table
- REPRODUCTION: Docs list
receipt_email; the schema only acceptscustomer_email, so the field is silently ignored - EVIDENCE: stripe-style-api.yaml, ChargeCreate:
customer_email: string - FIX: Rename to
customer_emailin the table and the sample
Next steps
- Apply E-01 and E-02 to every Node.js sample on the page
- Fix D-01, then re-run the Example and Consistency Hunters to confirm the score drops to 0
Connects securely to your tools. The creator never sees your data.
About this skill
Manual technical documentation review is slow, inconsistent, and prone to missing critical failures that break the developer experience. This skill runs a structured, bug-bounty-style audit to find broken code samples, API drift, and documentation dead ends before they reach your users.
What it does
- Multi-pass hunting executes specialized sweeps for code examples, schema drift, reference accuracy, and user journeys.
- Proof-based reporting requires every finding to include a specific location, reproduction steps, and evidence from a source of truth.
- Risk scoring calculates a leaderboard based on severity levels from Blocker to Informational.
- Code verification traces or executes snippets to catch syntax errors, missing imports, and mismatched outputs.
- Safety auditing identifies insecure defaults, exposed secrets, and destructive commands lacking warnings.
How it works
- Intake defines the audit scope, target files, and the source of truth like an OpenAPI schema or source repo.
- Execution runs independent hunter passes to identify defects without merging different bug categories.
- Triage assigns severity levels and points to every valid finding, filtering out style preferences or out-of-scope items.
- Reporting delivers a final ship/no-ship verdict, risk leaderboard, and paste-ready fixes.
Frameworks & tools
Works with any technical documentation format including Markdown, Docusaurus, and Sphinx. Integrates with OpenAPI/Swagger schemas, GraphQL definitions, and source code in any language for drift detection.
Why this beats prompting it yourself
Generic prompts often focus on grammar and tone while ignoring whether the code actually works. This skill enforces a "reproduce or it didn't happen" rule, ensuring every reported bug is a verified technical failure rather than a subjective opinion.
Use cases
- Pre-release audit of a new SDK or API version to ensure all samples match the updated schema.
- Quickstart stress-test to verify a first-time user can reach a "hello world" state without hidden prerequisites.
- Security sweep to catch hardcoded keys or dangerous copy-paste patterns in tutorials.
Known limitations
Findings are marked as unverified if no source of truth like a schema or repo is provided. Execution is restricted to sandboxed environments using placeholder credentials.
Behavior findings are marked Unverified when no source of truth (schema, repo, or CLI help) is provided. Code samples are only executed if your agent has an isolated sandbox; otherwise they are traced by reading. It doesn't judge product design, marketing copy, or translations.
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 21 days ago
Permissions
Read Files: reads the docs and source of truth (schema, code, CLI help) you point it at. Terminal / Shell: optional, only to run code samples in an isolated sandbox with test credentials. No network access or secrets required.