The Agent Skills Directory

[PROMPT_INJECTION]: Indirect prompt injection vulnerability surface. \n- Ingestion points: Untrusted data enters the context via external files including qualification-meeting-notes.md, discovery-notes.md, prospect-profile.md, and the user_input field. \n- Boundary markers: Absent. The skill does not employ delimiters (e.g., XML tags or triple quotes) or explicit instructions to the agent to disregard potential commands embedded within meeting notes. \n- Capability inventory: The skill's capabilities are restricted to generating and updating markdown files (qualification-scorecard.md, commercial-state.md). It does not possess network access, shell execution, or file system permissions outside of its operational files. \n- Sanitization: No filtering, validation, or escaping of input data is performed before it is processed by the agent.

commercial-qualification