The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/run_pacsomatic.py script utilizes subprocess.run to execute various system and scheduler commands, including java, bsub, sbatch, and qsub.- [COMMAND_EXECUTION]: The skill dynamically generates shell launch scripts tailored to the user's execution environment. These scripts are programmatically granted executable permissions using os.chmod(..., 0o755) and subsequently executed.- [REMOTE_CODE_EXECUTION]: The helper script includes functionality to clone a remote Git repository. While configured for a well-known bioinformatics pipeline by default, the repository URL is a user-adjustable parameter.- [PROMPT_INJECTION]: The skill processes user-provided inputs such as BAM file paths and sample IDs that are used to build configuration files and shell commands, creating an attack surface for indirect prompt injection.
Ingestion points: Command-line arguments for tumor/normal BAM paths, sample metadata, and repository URLs in scripts/run_pacsomatic.py.
Boundary markers: No specific delimiters or safety warnings are included in the instructions to separate untrusted data from processing logic.
Capability inventory: File system write access, file permission modification, repository cloning, and shell command execution.
Sanitization: Employs shlex.quote for shell command construction and uses a custom validator to prevent shell metacharacters in environment module commands.

pacsomatic