orbit-general

Pass

Audited by Gen Agent Trust Hub on May 7, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill aggregates data from a wide variety of third-party platforms to create a unified dashboard, which presents a surface for indirect prompt injection.
  • Ingestion points: The skill retrieves live data from at least 10 connectors including Slack, Notion, Gmail, GitHub, and Sentry as described in SKILL.md.
  • Boundary markers: The instructions do not provide clear delimiters or guidelines to separate the retrieved external data from the agent's core instructions.
  • Capability inventory: The skill produces interactive HTML output and is instructed to inject a specific JavaScript block from a template into the final rendered dashboard.
  • Sanitization: No sanitization or validation steps are defined for the content retrieved from external connectors, which could allow malicious instructions embedded in that data to be processed by the model.
  • [EXTERNAL_DOWNLOADS]: Fetches design assets from a well-known service to render the dashboard interface.
  • Evidence: Loads typography from Google's public font registry via a stylesheet import in example.html.
Audit Metadata
Risk Level
SAFE
Analyzed
May 7, 2026, 03:23 PM
Security Audit — agent-trust-hub — orbit-general