The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional markdown and platform configuration. It does not contain any executable scripts, binary files, persistence mechanisms, or hardcoded credentials.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it is designed to ingest and analyze untrusted data from external web applications and APIs. However, it incorporates strong mitigations by instructing the agent to treat target content as untrusted evidence that cannot override user instructions or expand the defined testing scope. Evidence: (1) Ingestion points: Web pages, source code, and API descriptions (SKILL.md, references/web-api-checks.md). (2) Boundary markers: Explicit instructions to ignore data-embedded commands and maintain authorization boundaries (SKILL.md, references/execution-safety.md). (3) Capability inventory: Access to HTTP clients, browser automation, and terminal tools for security testing. (4) Sanitization: Guidelines for redacting sensitive data and validating destination URLs before interaction (references/execution-safety.md).
[EXTERNAL_DOWNLOADS]: The skill contains references to established security methodology documentation from OWASP and PortSwigger. These are documented as external resources for educational purposes and do not involve the automated download or execution of remote code.

security-pentest