The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill manages proxy configuration files that store sensitive information, including usernames and passwords. It explicitly prompts the user for these credentials to populate residential proxy settings.
[COMMAND_EXECUTION]: The skill uses shell commands to perform system diagnostics, such as scanning listening ports with lsof and querying macOS network services via networksetup to inspect active proxy settings.
[PERSISTENCE]: The skill includes instructions to modify the user's shell configuration files (e.g., ~/.zshrc) to persistently set proxy environment variables across terminal sessions.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a potential attack surface by ingesting external configuration data (Clash profiles) which may originate from untrusted subscription URLs, combined with the capability to execute shell commands.
Ingestion points: The skill reads profiles.yaml and individual profile configuration files located in ~/Library/Application Support/io.github.clash-verge-rev.clash-verge-rev.
Boundary markers: No specific delimiters or validation logic is defined to sanitize or isolate data read from external configuration files.
Capability inventory: The agent has access to Bash for shell command execution and Write/Edit tools for modifying system files.
Sanitization: The instructions do not specify any sanitization or integrity validation of the content within the processed profile files before they are interpreted or used.

clash-doctor