The Agent Skills Directory

[COMMAND_EXECUTION]: The skill description specifies that it executes the brand-voice-scorer.py script by interpolating user-provided content directly into a shell command: brand-voice-scorer.py --brand {slug} --text "{translated_content}". This pattern is vulnerable to command injection if the translated_content contains shell-special characters such as semicolons, backticks, or subshell invocations, potentially allowing for arbitrary code execution in the context of the agent.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process text from untrusted external sources, including URLs and file paths. This creates a vulnerability surface where malicious instructions embedded in the source or translated content could be interpreted by the agent during the scoring and evaluation process. Ingestion points: original_content and translated_content provided as URLs or file paths. Boundary markers: None explicitly mentioned to separate untrusted data from instructions. Capability inventory: The skill has the ability to execute multiple Python scripts (language-router.py, eval-runner.py, brand-voice-scorer.py) and read local brand configuration files. Sanitization: No sanitization or validation of the ingested external content is described before it is passed to the scripts.

multilingual-score