to-android-probe
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data which could contain instructions meant to influence the agent's behavior.
- Ingestion points: The skill reads target Android application source code in step 3 and inspects JSON evidence files (chain, state, and UI data) in step 6.
- Capability inventory: The skill possesses the ability to generate and execute code on a mobile device through sibling skills, and has read access to the workspace file system.
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands are used when the agent ingests app source or evidence data.
- Sanitization: There is no mention of sanitization, filtering, or validation of the external content before it is processed by the agent.
- [DYNAMIC_EXECUTION]: The skill dynamically generates and executes scripts as its primary function.
- Evidence: The workflow involves generating a probe script using a sibling skill and executing it via a separate execution tool. While this is the intended purpose, it represents a high-privilege execution pattern.
Audit Metadata