The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Git commands for versioning and moving memory files. These commands are executed via subprocess.run using secure list-based arguments that prevent shell injection.
[DATA_EXPOSURE]: Analysis confirms the skill only accesses memory files within designated workspace roots. No attempts to access credentials, SSH keys, or other sensitive system paths were detected.
[REMOTE_CODE_EXECUTION]: The skill relies on local code and standard dependencies. It uses yaml.safe_load to process file metadata, which prevents arbitrary code execution during parsing.
[PROMPT_INJECTION]: While the skill processes user-controlled memory content, its operations are restricted to maintenance and structural analysis. It does not evaluate memory content as instructions, effectively mitigating standard prompt injection risks.

custodian