physics-quantum

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied problem descriptions to construct models, creating a surface for potential instruction injection within input data.\n
  • Ingestion points: User-provided problem descriptions are ingested into the context during the 'Define the system' phase (SKILL.md).\n
  • Boundary markers: Absent; there are no specific delimiters or instructions to ignore commands within user input.\n
  • Capability inventory: The skill is restricted to generating text-based artifacts and has no access to shell tools, network resources, or the local file system.\n
  • Sanitization: Absent; no validation or filtering of user-supplied descriptions is implemented.\n- [NO_CODE]: The skill consists entirely of natural language instructions and reasoning steps with no accompanying scripts or executable code.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 03:00 AM
Security Audit — agent-trust-hub — physics-quantum