The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to extract and process HTML comment markers from files to inform the agent's interactive behavior. This allows content from external files to influence the agent's instruction stream, which is a classic indirect prompt injection surface.
Ingestion points: File content containing HTML comment markers (referenced in SKILL.md description).
Boundary markers: None identified in the skill instructions to distinguish between data and instructions within comments.
Capability inventory: The skill specifies capabilities to interactively address comments and modify file content (removing markers).
Sanitization: No evidence of input validation or content sanitization for the extracted markers before they are processed by the agent.

process-review