The Agent Skills Directory

[SAFE]: The skill's primary function is to enhance privacy by processing sensitive data locally. It includes clear documentation and warnings regarding the privacy implications of using remote LLM endpoints.
[SAFE]: The agent_local.py script implements strict directory confinement. It uses os.path.realpath to resolve symbolic links and os.path.commonpath to validate that all file operations (read, list, search) remain within the user-authorized directory, effectively preventing directory traversal attacks.
[SAFE]: The codebase is authored using only the Python standard library (stdlib), avoiding risks associated with unverified third-party dependencies or malicious package updates.
[SAFE]: File access tools are read-only. The skill does not provide the local LLM with any capabilities to write files, execute shell commands, or perform network operations (other than communicating with the designated LLM API).
[SAFE]: The skill handles symbolic links and recursive directory structures cautiously; the confined_glob function explicitly skips directory links during recursive searches to prevent infinite loops and sandbox escapes.

qwen-delegate