The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from local requirement and technology selection documents, creating a potential vector for indirect prompt injection.\n
Ingestion points: Input is read from files matching requirements-{id}.md and technology-selection-{id}.md via the Read tool specified in the SKILL.md frontmatter.\n
Boundary markers: The instructions do not define delimiters, XML tags, or specific isolation protocols to distinguish between user-provided data and the agent's internal logic.\n
Capability inventory: The skill utilizes Read, Grep, and Glob tools and interfaces with a history-manager skill to track and update project phases.\n
Sanitization: The skill lacks explicit filtering, escaping, or validation steps for the content ingested from the iteration documents before processing it for technical review.

technical-review