memory-system
Pass
Audited by Gen Agent Trust Hub on Jul 1, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill facilitates persistent context management using local file operations (Read, Write, Grep, Glob) within the
.agents/memory/directory. It does not initiate network connections, download external resources, or execute arbitrary code. - [DATA_EXFILTRATION]: The skill provides proactive safety instructions in the 'What NOT to Save' section, specifically directing the agent to never record secrets, credentials, tokens, or private keys in the persistent memory files.
- [PROMPT_INJECTION]: The skill processes data from persistent local files, which creates a surface for indirect prompt injection if those files are modified by untrusted entities. The skill mitigates this by instructing the agent to distill insights rather than store verbatim conversation transcripts.
- Ingestion points: Reads from
.agents/memory/MEMORY.mdand related topic files during session start and recall operations. - Boundary markers: Recalled context is applied silently without explicit delimiter-based isolation from the current session context.
- Capability inventory: Read, Write, Grep, and Glob tools used to manage the knowledge base.
- Sanitization: Relies on the agent's internal reasoning to filter and summarize information during the 'Save' operation.
Audit Metadata