The Agent Skills Directory

Data Processing Considerations: The skill is designed to ingest and process meeting transcripts and notes from external connectors (meeting_notes) and user-provided text. This is an intended functionality for sales analysis.
Indirect Prompt Injection Surface: Like any skill that processes external content, there is a theoretical surface where malicious instructions could be embedded within transcripts to influence the agent's output. The skill relies on standard model guardrails and its own evidence-led instructions to mitigate this.
Ingestion points: Meeting transcripts, call notes, and participant context retrieved via the meeting_notes connector tools and user-pasted content.
Boundary markers: The instructions focus on extraction of specific evidence but do not explicitly define technical boundary markers for the ingested data.
Capability inventory: The skill uses retrieval tools for searching and fetching call data; it does not utilize sensitive system operations such as arbitrary code execution or direct file system writes.
Sanitization: The skill emphasizes being 'evidence-led' and using short quotes, which serves as a functional filter for relevant content.
Vendor-Internal Resource Usage: The skill references internal ecosystem components like sales:user-context for preflight configuration and context resolution, which is a standard pattern for modular agent skills.

get-rep-call-feedback