How to use the API Security Testing agent skill

What the skill makes the agent do

The skill gives the agent a step-by-step methodology for finding security weaknesses in APIs, the interfaces apps use to talk to servers. It targets REST, GraphQL and gRPC APIs and uses the OWASP API Security Top 10 as its testing framework.

The methodology covers five areas: discovering endpoints, testing logins and tokens, checking whether users can reach data or actions they should not, probing inputs for injection flaws, and checking whether responses leak too much information.

• Discovery: importing API documentation, exporting a Burp sitemap, sending a GraphQL introspection query, and enumerating hidden or debug endpoints
• Authentication: JWT analysis, OAuth 2.0 checks and API key handling
• Authorization: BOLA (IDOR) and BFLA tests, mass assignment and HTTP method testing
• Input validation: SQL, NoSQL, SSRF and XXE checks, plus rate limiting validation
• Data exposure: verbose responses, error messages, pagination enumeration and debug endpoints

Install the skill from the ZIP

Download the ZIP for this skill from this page. You can install it in one of two ways. For a personal setup, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder. Alternatively, in the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.

After unzipping, the skill folder should contain SKILL.md, scripts/agent.py and references/api-reference.md.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload

Check what you need before you start

The skill assumes you already have authorization and a working test setup. Confirm these before you ask the agent to test anything.

Python 3 with the requests library is needed for scripts/agent.py. Burp Suite Professional and Postman or Insomnia are recommended and referenced in the workflow. You also need valid test credentials at multiple privilege levels (unauthenticated, standard user and admin are the levels the skill names) and written authorization for the target. hashcat is needed only if you perform the JWT secret brute-force step. The Burp Autorize extension is optional and automates BOLA checks.

• Python 3 and the requests library (requests >=2.28 per the reference file)
• Burp Suite Professional (recommended)
• Postman or Insomnia (recommended)
• Test credentials at multiple privilege levels
• Written authorization for the target
• hashcat, only for JWT secret brute-forcing
• Burp Autorize extension (optional)

Trigger the skill with a natural request

You do not need special syntax. The skill activates for requests about API security testing, REST API pentests, GraphQL security assessments or API vulnerability testing. State the target, the scope and the credentials you are providing, and say that you have authorization.

A useful request names the API base URL, the test scope and which tokens are available. For example, you might write: "I have written authorization to test the staging API at the base URL below. Use the API security testing skill. I have a standard user token and an admin token. Check for BOLA and BFLA on the user and order endpoints, and check whether GraphQL introspection is enabled."

Run the included script for a BOLA check

The skill includes scripts/agent.py, which automates a basic check for broken object-level authorization by comparing responses for two user IDs. The reference file shows the command-line form below. Arguments include --base-url (required), --token, --low-priv-token for BFLA testing, --login-endpoint for the rate limiting test (default /api/auth/login), --graphql for introspection disclosure testing, and --output to write a report.

Only run it against systems you are authorized to test.

Read the results and follow up

Treat the agent's output as a list of leads, not a verdict. A flagged BOLA response still needs a manual check in Burp Suite or Postman to confirm the data really belongs to another user. The workflow asks the tester to replay requests across sessions, so the results should be reproducible before you report them.

Ask the agent to walk through a single finding in detail if you want to confirm it. Keep the test notes and the report file with your engagement records.

Safety notes for scripts and network use

This skill runs scripts and uses the network. The script sends HTTP requests to the base URL you give it, so point it only at systems you are authorized to test, and never at production systems that handle real payments without safeguards. The skill itself says not to use it against APIs without written authorization, and not for load testing or denial-of-service testing unless explicitly scoped.

Some workflow steps are deliberately disruptive. The skill's rate limiting step sends 100+ rapid requests to login, password reset and OTP endpoints, which can trigger lockouts. Injection and brute-force payloads can also cause damage. Agree the scope and timing in advance, and read scripts/agent.py before running it.

• Use only against authorized targets, ideally staging or a lab
• Avoid load and denial-of-service testing unless explicitly scoped
• Expect lockouts from the rate limiting step
• Review scripts/agent.py before execution

Tips for better results

Give the agent complete documentation where you have it, such as an OpenAPI spec, a GraphQL schema or a Postman collection, because the discovery step starts there. Provide tokens for each privilege level, since BOLA and BFLA checks need at least two identities to compare.

Ask for one area at a time, such as authentication or authorization, so findings stay easy to verify. If you are a developer, run the checks against your own staging API before release.

• Supply an API spec or Postman collection to speed up discovery
• Provide at least two identities for BOLA comparisons
• Test one category at a time
• Use a deliberately vulnerable lab to practise OWASP API Top 10 issues