java-refactor-engram-contract
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill defines a protocol for handling data from a memory system, representing a surface for indirect prompt injection. Analysis: 1. Ingestion points: Artifacts are retrieved via
mem_get_observation(SKILL.md). 2. Boundary markers: The skill requires explicit validation ofprojectandrun_idand enforces strict namespace isolation for topic keys. 3. Capability inventory: Access is limited to platform-specific memory tools (mem_save,mem_search,mem_get_observation). 4. Sanitization: Instructions mandate compact evidence and explicitly forbid storing raw source code or logs. - [SAFE]: The skill implements strong security hygiene by preventing session-based inference and enforcing explicit decision gates before data access.
Audit Metadata