The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The documentation describes installing the Pixi package manager using the command curl -fsSL https://pixi.sh/install.sh | bash. This pattern involves downloading and executing a remote script in a single step, which is a common but risky installation method that bypasses traditional package integrity checks.
[EXTERNAL_DOWNLOADS]: Fetches configuration and installation scripts from the official Pixi package manager domain (pixi.sh).
[CREDENTIALS_UNSAFE]: The connection helper examples/wrds_conn.py reads sensitive database credentials from the ~/.pgpass file. While utilizing a dedicated credentials file is a standard security practice for PostgreSQL, the programmatic extraction of these secrets for URI construction involves handling plaintext credentials from the filesystem.
[INDIRECT_PROMPT_INJECTION]: Slurm job templates use sed to extract tasks from an external file (items.txt) and pass them as command-line arguments to Python scripts. Ingestion points: Task items are read from a file specified by the TASK_LIST environment variable. Boundary markers: None. Capability inventory: The skill facilitates shell command execution and environment manipulation. Sanitization: None; the data is passed directly as a command-line argument, creating a surface for potentially malicious content in the task file to influence execution.

hpc