memory-management
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves ingesting data from external sources (chat, email, calendar, documents) via the
/productivity:startcommand and user interactions to populate persistent memory files (CLAUDE.mdand thememory/directory). This ingestion process lacks sanitization or boundary markers, creating a surface for indirect prompt injection where malicious instructions embedded in external content could be stored and subsequently acted upon by the agent when it recalls that memory. \n - Ingestion points:
CLAUDE.md,memory/glossary.md,memory/people/,memory/projects/, and external data scanned by the/productivity:startcommand. \n - Boundary markers: None. The instructions do not specify the use of delimiters or 'ignore' instructions for the stored data. \n
- Capability inventory: The skill possesses file read and write capabilities across the local directory to manage context. \n
- Sanitization: None. The skill does not provide instructions for validating or escaping content retrieved from external inputs before committing it to memory.
Audit Metadata