memory
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands using repository paths provided by the user (e.g.,
sessions memory mine --repo <path>). This presents a command injection risk if the agent does not sanitize the user-supplied path before passing it to the CLI tool. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and triages data from past AI session history, which may contain adversarial content. Ingestion points: Data is pulled from the output of
sessions memory mine. Boundary markers: No explicit delimiters or 'ignore' instructions are used during processing. Capability inventory: The skill can write durable instructions to the agent's memory store viasessions memory approve. Sanitization: The skill includes manual rubric checks for prompt-injection patterns and relies on backend tool validation to refuse malicious content. - [DATA_EXPOSURE]: The skill explicitly accesses and reads the local file
~/.local/share/sessions/groups.jsonto manage project group assignments.
Audit Metadata