memory-engineering
Pass
Audited by Gen Agent Trust Hub on Aug 23, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill includes four Python scripts for memory analysis and auditing. These scripts utilize the Python standard library (argparse, json, re, os) to perform deterministic analysis on provided data. There is no evidence of arbitrary shell execution or unsafe command interpolation.
- [DATA_EXFILTRATION]: While the skill processes memory directories (e.g.,
~/.claude/memory), all analysis is performed locally. The scripts do not include network-capable modules or protocols, and all outputs are directed to the standard output for user review. - [SAFE]: The skill emphasizes security and data integrity by requiring that contradictions be surfaced to a human rather than automatically merged, and by enforcing strict forgetting policies to prevent uncontrolled state growth.
- [SAFE]: All external references provided in the documentation target well-known research repositories (arXiv) and trusted GitHub organizations (Anthropic, Microsoft). These links are for informational purposes and do not trigger automated downloads or execution.
Audit Metadata