The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from CRM records and interaction histories, creating a vulnerability surface for indirect prompt injection.
Ingestion points: As described in references/task-method.md, the skill ingests "active-opportunity roster", "stage history", "earlier communication", and "Motion/cell narratives" from external systems.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore or isolate instructions potentially embedded in the CRM fields or communication transcripts.
Capability inventory: The skill has the capability to execute a local Python script for state management and can transition to other agent skills based on the analyzed data.
Sanitization: No sanitization, validation, or filtering procedures for the ingested data are defined in the skill logic.
[COMMAND_EXECUTION]: The skill utilizes a local utility script to manage its persistence and internal state tracking.
Evidence: references/briefing-output-template.md specifies the use of ../../shared/scripts/workspace_state.py to store detector results and observation history.

signals