android-benchmark-comparison
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and interpret external data including raw benchmark results, traces, and build identities, which creates a potential surface for indirect prompt injection.
- Ingestion points: According to the procedures in
SKILL.md, the agent processes 'raw results', 'traces', 'metric definitions', and 'exact build identity'. - Boundary markers: The instructions do not define specific delimiters or guardrails to prevent the agent from being influenced by instructions that might be embedded within trace files or benchmark metadata.
- Capability inventory: The skill references the execution of 'device-specific commands' (procedure #2) and 'CPU masks', as well as the ability to 'discover the device topology' (procedure #4).
- Sanitization: There are no instructions provided for sanitizing, validating, or escaping the external performance data before processing.
Audit Metadata