The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill instructions direct the agent to ingest content from an external file (.claude/product-marketing-context.md) which could contain malicious instructions.
Ingestion points: .claude/product-marketing-context.md (referenced in SKILL.md in the Initial Assessment section).
Boundary markers: Absent. The agent is simply told to "read it before asking questions."
Capability inventory: None. The skill contains only markdown instructions and no functional code, command execution, or network access capabilities.
Sanitization: Absent.
Assessment: While the injection surface exists, the lack of side-effect capabilities (read/write/execute) limits the impact to potential bias in the generated text recommendations.

onboarding-cro