The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data in the form of plain-text requests to triage and route project tasks.
Ingestion points: Ingests user-provided requests in SKILL.md to classify intent and consequence.
Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions embedded within the user requests.
Capability inventory: The skill can trigger rollbacks (via the setup-hosting skill), modify project pieces and labels (via the setup-ai-build-kit skill), and update documentation like the masterplan or CHANGELOG.md.
Sanitization: There is no mention of sanitizing or escaping the user-provided text before it is used to influence the agent's routing logic.

change-triage