The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes documents and communication history provided by external project participants, creating an entry point for indirect prompt injection.
Ingestion points: The skill is designed to read and integrate feedback from external CRD files, communication logs (emails/chats), and project documentation to update requirements (documented in SKILL.md Sections 1 and 4).
Boundary markers: While the skill enforces the use of a structured Markdown template (assets/cr-template.md) to organize data, it lacks explicit instructions or technical controls to sanitize or ignore instructions that might be embedded within the external text.
Capability inventory: The skill manages file operations (reading and writing Markdown files) and utilizes interactive tools to resolve technical gaps or seek user confirmation.
Sanitization: There are no documented procedures for the validation or sanitization of ingested content to ensure it does not influence the agent's logic.

crd