The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes user-supplied architecture documents and source code, which are untrusted inputs. Malicious instructions hidden within these inputs could potentially influence the agent's threat assessment or bias the generated report.\n- Ingestion points: Step 1 of the Process instructs the agent to read architecture docs or code (SKILL.md).\n- Boundary markers: The instructions lack explicit delimitation or warnings to treat processed content strictly as data.\n- Capability inventory: The skill is limited to document generation and lacks file-write, network, or code execution permissions.\n- Sanitization: No input validation or instruction filtering is performed on the ingested content.

threat-modeling