The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md direct the agent to execute a local Node.js script located at scripts/generate_report.js to process JSON audit data into a Word document report.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to analyze untrusted data (advertisement materials) provided by users, which creates a potential surface for injection attacks designed to influence the audit outcome.
Ingestion points: Advertisement text and material descriptions provided by the user in the agent's chat context.
Boundary markers: None explicitly defined in the skill instructions to separate the agent's instructions from the untrusted advertisement content.
Capability inventory: The skill has the capability to write files to the local file system and execute local scripts.
Sanitization: No explicit sanitization or filtering of the user-provided content is performed before it is analyzed or passed to the report generation script.
[DYNAMIC_EXECUTION]: The scripts/generate_report.js file utilizes child_process.execSync to run npm config get prefix at runtime. This is used to dynamically determine the path to the global node_modules directory to resolve the docx dependency.

ad-compliance-review