The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/run_muse.py uses dynamic path manipulation to import code. It modifies the Python path via sys.path.insert(0, ...) using a calculated path (../../cline-runner/scripts) located two levels above its own directory. This allows the execution of modules from an external location not contained within the skill's package structure.
[COMMAND_EXECUTION]: The skill's primary function is to invoke the cline CLI tool to run prompts, which involves executing shell commands through a runner wrapper.
[EXTERNAL_DOWNLOADS]: The SKILL.md file instructs users to perform a global installation of the cline tool from a public package registry (npm install -g cline).
[DATA_EXFILTRATION]: The skill's security model explicitly notes that prompt text, metadata, and files read during execution are transmitted to external service providers (e.g., OpenRouter).
[PROMPT_INJECTION]: The skill acts as an interface for processing external data and user prompts, creating a surface for indirect injection.
Ingestion points: User-provided prompts and prompt files referenced in scripts/run_muse.py.
Boundary markers: None; the wrapper script does not implement delimiters or safety instructions to separate data from commands.
Capability inventory: The skill executes the cline tool, which has broad capabilities including file system access and network operations.
Sanitization: No input validation or sanitization logic is present in the wrapper script.

muse-runner