memory-file-maintenance
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided context files such as
MEMORY.mdorCLAUDE.md. While these files contain natural language instructions, the skill serves as a transformation and review tool rather than an execution environment. - Ingestion points: User-supplied content from memory/context files provided during interaction.
- Boundary markers: The skill relies on natural language prompting to distinguish between the instructions to be reviewed and the review task itself.
- Capability inventory: The skill does not invoke any shell commands, network operations, or privileged file system access.
- Sanitization: The framework explicitly mandates a 'Privacy re-check' to identify and remove credentials, financial data, health information, or other personal data that should not be stored in AI context files.
- [CREDENTIALS_UNSAFE]: The skill mentions sensitive terms like 'credentials' and 'secrets' exclusively within the context of a safety mechanism intended to help the user identify and remove such items from their memory files.
Audit Metadata