mobile-instinct-v2
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze development session data, including code changes and problem-solving contexts, to extract reusable patterns. This creates a potential surface for indirect prompt injection if malicious instructions are embedded within the code or context being observed.
- Ingestion points: The skill monitors code changes, problem contexts, and manual inputs provided via the
/instinct-observecommand. - Boundary markers: The instructions do not specify the use of delimiters or boundary markers to distinguish between untrusted session data and system instructions.
- Capability inventory: The skill utilizes session hooks (
hooks/instinct-hooks.json) and pattern extractors (agents/mobile-pattern-extractor.md) to process and persist data across multiple sessions. - Sanitization: There is no evidence of sanitization, filtering, or validation of the observed data before it is incorporated into the agent's learned 'instincts'.
Audit Metadata