The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes the entire codebase of an application for its assessment, which creates an attack surface for indirect prompt injection. A malicious file within the scanned project could contain instructions that attempt to influence the agent's behavior during the audit.
Ingestion points: SKILL.md instructs the agent to "Use the entire codebase as the assessment scope."
Boundary markers: The instructions do not define specific delimiters or "ignore instructions" warnings to help the agent distinguish between code meant for analysis and potential instructions embedded within that code.
Capability inventory: The agent is tasked with writing a comprehensive HTML report to the local file system (codepulse-full/reports/).
Sanitization: No specific sanitization or escaping instructions are provided for when the agent interpolates discovered code snippets or vulnerability findings into the HTML report template, potentially allowing for stored cross-site scripting (XSS) in the generated report if the scanned code contains malicious payloads.

codepulse-full