signals-scout-data-warehouse

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from external data source errors and user activity logs, creating an attack surface for indirect prompt injection.
  • Ingestion points: The skill ingests untrusted text via the latest_error field from system.source_schemas (referenced in SKILL.md), user-authored SQL strings from the query_log table, and view definitions from system.data_modeling_views.
  • Boundary markers: The instructions in SKILL.md explicitly direct the agent to treat this external text strictly as data and never as instructions, specifically advising to quote them as short snippets.
  • Capability inventory: The agent can execute SQL queries (execute-sql), write to its persistent scratchpad (scout-scratchpad-remember), and author/edit reports (emit_report, edit_report) that are surfaced to human reviewers.
  • Sanitization: The skill guidelines suggest truncating long error messages and dropping payload echoes when including external data in reports.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 03:59 PM
Security Audit — agent-trust-hub — signals-scout-data-warehouse