personal-crm

Pass

Audited by Gen Agent Trust Hub on Aug 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest data from external meeting recorders (e.g., Granola, Fireflies, Fathom) and process transcripts into dossier notes. This creates a surface where malicious instructions embedded in a meeting transcript could influence the agent's behavior during the normalization and extraction process.
  • Ingestion points: Meeting notes and verbatim transcripts are pulled from external meeting-recorder MCPs and written to the meetings/ directory (SKILL.md).
  • Boundary markers: The skill defines a --- timeline --- separator for dossier files and a specific dated-bullet format, which helps structure data but does not provide security isolation for untrusted content.
  • Capability inventory: The agent uses the write() tool to create and update dossier files and interacts with various registered MCP servers for data retrieval.
  • Sanitization: The instructions emphasize verbatim recording of transcripts and extraction of entities but do not specify sanitization or filtering of the external content to prevent command or prompt injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 30, 2026, 04:21 PM
Security Audit — agent-trust-hub — personal-crm