The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external code snippets, which presents a surface for indirect prompt injection where malicious instructions could be embedded in the code being reviewed. However, the skill includes strict instructions forbidding dynamic execution and mandates a structured reporting format, which mitigates the risk of the agent following instructions contained within the analyzed data.
Ingestion points: The code input defined in SKILL.md is the primary entry point for untrusted data.
Boundary markers: The instructions use structured tags like <critical_constraints> and <owasp_checklist> to define the operational scope.
Capability inventory: The skill has filesystem:read permissions and includes a Python script scripts/validate.py that reads a local file to verify report formatting. It lacks network access or code execution capabilities.
Sanitization: The skill relies on LLM-level constraints to prevent execution of input data, explicitly stating 'NO execution/dynamic analysis' in its constraints.
[COMMAND_EXECUTION]: The provided script scripts/validate.py is a benign utility used for validating the format of generated security reports. It does not take user-controlled arguments or perform dangerous shell operations.

ask-owasp-security-review