The Agent Skills Directory

[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The buildProxy function in runtime/lib/onboarding.js clones a remote Git repository from an unverified source (https://github.com/router-for-me/CLIProxyAPI.git) and executes code from it by running go test and go build during the installation process.
[PERSISTENCE_MECHANISMS]: The parable.sh script modifies shell profile files such as .bashrc, .zshrc, and .profile to add its installation directory to the system PATH, ensuring the tool remains available across user sessions.
[DYNAMIC_EXECUTION]: The verify command in scripts/parable.py uses subprocess.run with shell=True to execute arbitrary shell commands defined in the user's parable.toml configuration file.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as repository notes, plans, and routing instructions (ingestion points in SKILL.md) without boundary markers or sanitization. This data is interpolated into prompts that drive powerful capabilities including shell command execution and remote repository cloning (capability inventory in scripts/parable.py and onboarding.js).
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill's primary script (scripts/parable.py) contains an error message that instructs users to execute a remote script via curl | bash, which is an insecure practice that can lead to remote code execution.

parable