catalyst-calendar

Pass

Audited by Gen Agent Trust Hub on Oct 11, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources via WebSearch and WebFetch to identify catalysts (Step 2).
  • Ingestion points: External websites, regulatory dockets, and news sources accessed via web tools.
  • Boundary markers: The instructions explicitly mandate the use of primary source overrides (issuers, regulators) and structured "Event Lenses" for extraction, which helps constrain the AI's interpretation of external content.
  • Capability inventory: The skill has the ability to write to files (calendar.xlsx) and execute a local script (recalc.py).
  • Sanitization: The skill relies on source verification and strict schema matching (event categories and materiality scoring) rather than programmatic sanitization.
  • [COMMAND_EXECUTION]: The skill invokes a local script (python .agents/skills/xlsx/scripts/recalc.py) to recalculate the generated Excel workbook. This is a legitimate use of internal functionality designed to support the skill's primary purpose of calendar generation.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 11, 2026, 01:05 AM