The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from external sources during the reconnaissance and vulnerability assessment phases, which creates a potential surface for indirect prompt injection.\n- Ingestion points: Processes external data such as OSINT, DNS records, public repositories, port scan results, and service banners as defined in references/recon_enumeration.md.\n- Boundary markers: The skill requires adherence to a signed Statement of Work (SOW) and Rules of Engagement (ROE), and instructs the analyst to manually validate each finding, providing a procedural boundary against automated obedience to embedded data.\n- Capability inventory: The skill describes workflows for reconnaissance, PoC development, and attack-path chaining, assuming access to standard offensive security tools.\n- Sanitization: The methodology relies on human-in-the-loop manual verification and repeatable steps for findings; it does not specify technical sanitization or filtering of external input data within the agent instructions.

offensive-security-analyst