trailmark-finding-triage

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill defines a methodical process for ingesting external data such as SARIF files and report excerpts, which constitutes a surface for indirect prompt injection. However, the instructions mitigate this by requiring strict normalization and instructing the agent to verify all claims through graph evidence rather than trusting the input text. \n
  • Ingestion points: External findings are accepted as input in Step 1 and handled according to references/input-normalization.md. \n
  • Boundary markers: The skill explicitly instructs the agent to "Extract the claimed impact but do not assume it is true" and provides templates to separate candidate claims from bound graph nodes. \n
  • Capability inventory: The skill utilizes the trailmark Python library and standard file system tools for analysis. \n
  • Sanitization: The requirement to produce a structured evidence packet based on graph queries effectively isolates the agent's logic from potential adversarial instructions in the findings.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 03:49 AM
Security Audit — agent-trust-hub — trailmark-finding-triage