memory
Pass
Audited by Gen Agent Trust Hub on Jul 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a vendor-specific command-line interface,
cli-jaw, to manage persistent memory and search chat history. These operations are performed on local files located in the{{JAW_HOME}}/memory/directory and local SQLite databases. - [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by design, as it allows the agent to retrieve and act upon previously stored information from past sessions or chat logs.
- Ingestion points: Historical data is ingested from memory files (
cli-jaw memory read) and chat databases (cli-jaw chat search). - Boundary markers: The instructions lack requirements for the agent to use specific delimiters or sanitization protocols when processing retrieved memories.
- Capability inventory: The agent has write access to memory storage via
cli-jaw memory save. - Sanitization: There is no evidence of filtering or validation for the content stored in or retrieved from the memory system.
Audit Metadata