api-security-testing

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly asks for API tokens/credentials and shows them embedded in the agent instruction block, and it produces "exact" proof-of-concept requests that would include those tokens verbatim (e.g., Authorization headers), so the LLM would need to handle/output secrets directly.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). SKILL.md describes an API security testing workflow where the LLM runtime ingests user-supplied instruction/context (including tokens and scope) and also parses specified target specs (OpenAPI/Postman/GraphQL schemas) or reads user-provided workspace files, which are free text supplied by the user/attacker, creating indirect prompt-injection exposure.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly supports pulling a live Postman collection at runtime via "-t postman://", causing the agent to fetch external collection content that directly controls which endpoints and actions the agent will perform (postman://).