mesh-memory
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides tools to store and retrieve user-provided notes, worklogs, and decisions which are then injected into the agent context. This creates a surface for indirect prompt injection attacks where stored instructions could override agent behavior.\n
- Ingestion points: The
mesh_addandmesh_updatetools inSKILL.mdare used to persist untrusted external data.\n - Boundary markers: While
SKILL.mdincludes a warning that retrieved content is 'untrusted historical context', it does not provide technical delimiters for the agent to distinguish this data from primary instructions.\n - Capability inventory: Tools in
SKILL.mdprovide capabilities for storing, searching, and retrieving documents via a semantic database.\n - Sanitization: There is no evidence of input validation or output sanitization for the stored text content within the skill definition in
SKILL.md.
Audit Metadata