The Agent Skills Directory

[COMMAND_EXECUTION]: The skill scaffolds Python classes that execute external CLI tools. For 'agent-runner' providers, it generates code using helper functions like run_cmd and _invoke_cli_agent to launch subprocesses. The logic includes quality gates that verify the use of shlex.split for argument parsing to mitigate shell injection risks.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to add installation steps to action.yml, which may involve fetching CLI tools via npm or vendor-provided installation scripts. These are standard procedures for integrating third-party LLM provider agents.
[PROMPT_INJECTION]: The skill takes several user inputs and interpolates them into code and configuration files. Ingestion points: provider_id, api_url, cli_bin, and install_command from the inputs section. Boundary markers: None present in the generated templates. Capability inventory: The skill uses Write and Edit tools to modify scripts/reviewer.py and action.yml. Sanitization: Quality gates verify that generated code uses shlex.split for command arguments.

add-provider