The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the installation of the alibi and alibi-detect libraries from the Python Package Index (PyPI). These are well-known open-source interpretability and monitoring tools developed by Seldon.- [COMMAND_EXECUTION]: The documentation provides command-line instructions for environment setup using pip install.- [PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface through a path traversal vulnerability in the provided logging code template in SKILL.md. The explain_and_log function uses the instance_id parameter directly to construct a file system path without sanitization or validation.
Ingestion points: The instance_id argument in the explain_and_log Python function (Step 6).
Boundary markers: Absent. The code does not use delimiters or instructions to ignore embedded control characters in the identifier.
Capability inventory: The skill uses pathlib.Path.write_text to perform file system write operations.
Sanitization: Absent. The provided example does not demonstrate filename validation or path normalization, allowing an attacker-controlled identifier (e.g., ../../etc/passwd) to potentially overwrite system files.

alibi-explainability