probe-debugging
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external firmware files and hardware states, which presents an attack surface for instructions embedded in data. • Ingestion points: ELF files, register values, and stack memory contents as noted in SKILL.md. • Boundary markers: No specific delimiters for untrusted data are defined in the instructions. • Capability inventory: The skill guides the use of GDB and OpenOCD for target control, memory writes, and register modifications in SKILL.md and references/workflow.md. • Sanitization: The instructions do not specify methods for filtering or escaping data retrieved from the hardware targets.
Audit Metadata