The Agent Skills Directory

[PROMPT_INJECTION]: The skill retrieves and summarizes external AI responses via the list_responses tool, creating a surface for indirect prompt injection. Malicious instructions within monitored content could potentially influence the agent's analysis.
Ingestion points: AI response data retrieved from the Promptwatch API.
Boundary markers: The skill lacks explicit markers to distinguish untrusted content from system instructions.
Capability inventory: The workflow includes sampling negative responses and summarizing sentiment drivers.
Sanitization: No explicit sanitization or filtering of external data is mentioned.
[EXTERNAL_DOWNLOADS]: The skill requires connecting to an external MCP server at https://server.promptwatch.com/mcp, which is hosted on the vendor's domain.

sentiment-monitor