gtm-daily-prospecting

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted data from an external CRM, specifically call and meeting transcripts, notes, and task subjects (Steps 1 and 3). This content is directly interpolated into the agent's context to generate personalized call openers and discovery questions.
  • Ingestion points: External CRM data including task notes, call transcripts, and meeting outcomes identified in Step 1 and Step 3.
  • Boundary markers: The instructions do not define clear delimiters (e.g., XML tags) or "ignore embedded instructions" markers for the external data being processed.
  • Capability inventory: The skill allows for querying a CRM and generating artifacts (Markdown, HTML, and PDF files).
  • Sanitization: There are no instructions for the agent to sanitize or validate the content retrieved from the CRM before processing it.
  • [DYNAMIC_EXECUTION]: The 'Artifact Generation' section for PDF output indicates the use of 'Python + reportlab' for rendering documents. This involves the dynamic generation and execution of scripts based on data gathered during the workflow.
  • [DATA_EXFILTRATION]: The skill instructions require the agent to load a configuration file (profiles/client-profile.md) located using a relative path that navigates outside the skill's root directory (../../). This cross-directory access relies on the agent's environment allowing file reads across the broader project structure.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 04:19 PM
Security Audit — agent-trust-hub — gtm-daily-prospecting