signals-scout-csp-violations

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes '$csp_violation' events which are client-reported and potentially attacker-controlled.
  • Ingestion points: Telemetry data including '$csp_blocked_url', '$csp_document_url', and '$csp_original_policy' are ingested via SQL queries in 'SKILL.md'.
  • Boundary markers: The documentation includes a specific section ('Read the policy from code, and reconcile') that warns the agent that reports can be forged and should not be trusted in isolation.
  • Capability inventory: The skill uses 'execute-sql' for data ingestion and the 'gh' CLI to read repository contents; findings are reported via 'emit_report'.
  • Sanitization: The agent is explicitly instructed to reconcile telemetry findings with trusted source code artifacts on GitHub before recommending any policy changes, which acts as a validation step against malicious data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 01:03 AM
Security Audit — agent-trust-hub — signals-scout-csp-violations