The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script scripts/clash_tun_doctor.py uses subprocess.run to manage the lifecycle of the Clash Verge application on macOS.
Evidence: It executes osascript to quit the application gracefully and uses the open command to restart it after configuration changes.
Evidence: It utilizes pgrep to verify the process state during repair operations.
[SAFE]: The skill is implemented using only the Python standard library and does not download or execute code from external sources.
Evidence: The script uses http.client and socket for local Unix Socket communication with the Mihomo API rather than external network requests.
[FILE_SYSTEM_ACCESS]: The skill reads and modifies application-specific configuration files and logs.
Evidence: It accesses ~/Library/Application Support/io.github.clash-verge-rev.clash-verge-rev to read logs and edit config.yaml and profile rules.
Evidence: It implements a mandatory backup system that creates timestamped copies of all modified files in a dedicated backup directory before applying changes.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from network logs which could potentially contain malicious payloads.
Ingestion points: The tool reads service_latest.log and active connection data from the local Mihomo API.
Boundary markers: The script does not use explicit boundary markers when outputting log data to the agent context.
Capability inventory: The skill has the ability to stop/start processes and write to the filesystem.
Sanitization: The script uses strict regular expressions (HOSTNAME_RE) to validate and normalize hostnames extracted from logs before processing them.

lovstudio-clash-tun-doctor