daily-task-prep

Pass

Audited by Gen Agent Trust Hub on Apr 15, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it retrieves and processes untrusted data from meeting attendee context and calendar entries.
  • Ingestion points: Untrusted data enters the context via get_page and get_timeline when loading attendee brain pages and recent timeline entries in Phase 1 and Phase 2.
  • Boundary markers: The prompt template does not use specific delimiters or instructions to treat external data as untrusted or to ignore embedded instructions.
  • Capability inventory: The skill is restricted to read-only operations (search, query, get_page, list_pages, get_timeline) and is explicitly marked as non-mutating, which mitigates the impact of potential injection.
  • Sanitization: No sanitization or validation of retrieved content is performed before it is interpolated into the final briefing output.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 15, 2026, 06:37 AM