The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute local scripts such as preflight.sh, build_uv_env.sh, and download_startup.sh, as well as the complexa CLI tool. These executions are intended for environment verification, dependency installation, and model management within the project workspace.
[EXTERNAL_DOWNLOADS]: Through the complexa download command, the skill fetches large model weights from established and well-known services, including NVIDIA NGC, GitHub, Hugging Face, and AWS. These operations are necessary for obtaining the required checkpoints for protein design pipelines.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied directory paths for local code and data storage, interpolating these values into shell commands and the .env file.
Ingestion points: Path variables LOCAL_CODE_PATH and LOCAL_DATA_PATH provided by the user in SKILL.md.
Boundary markers: The skill does not use explicit boundary delimiters when interpolating these paths into commands.
Capability inventory: The skill utilizes Bash for command execution and Write for environment file modification.
Sanitization: There is no evidence of specific input validation or sanitization for the user-provided paths.
[DYNAMIC_EXECUTION]: The skill dynamically generates a run_manifest.json setup artifact using a Python script and modifies the .env file based on the user's choice of runtime (UV or Docker). These are standard configuration procedures for local project initialization.

complexa-setup