daily-sales-report

Pass

Audited by Gen Agent Trust Hub on Sep 25, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from multiple sources, which presents an attack surface for indirect prompt injection.
  • Ingestion points: The skill ingests data through MCP tool connections, CSV file exports, manual user pastes, and external web content fetched during company research (documented in SKILL.md and references/house-rules.md).
  • Boundary markers: The skill explicitly mitigates these risks via references/agent-security.md, which instructs the agent to treat all retrieved content as data, never as instruction, and to quote rather than summarize injection attempts.
  • Capability inventory: The agent is authorized to write to internal state files in the .agents/ directory, perform web fetches for research, and render visual pulse boards via platform-specific tools.
  • Sanitization: The skill mandates HTML-escaping for all interpolated values and forbids the use of dangerous rendering methods like innerHTML (references/agent-security.md).
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to fetch and analyze company websites and public competitor data to build a product context. These network operations are intended for research purposes and are governed by the security rules mentioned above.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 25, 2026, 04:06 PM