The Agent Skills Directory

[SAFE]: The skill logic in src/index.js and the supporting Python scripts perform expected prompt-engineering tasks without any detected malicious patterns. The code handles user input through a normalization and safety-check process.
[COMMAND_EXECUTION]: Local Python scripts (scripts/validate_output.py, scripts/validate_creative_artifacts.py) utilize subprocess.run to invoke internal validation utilities. These operations are restricted to the local environment and are used to ensure artifact quality, presenting no risk of unauthorized remote command execution.
[PROMPT_INJECTION]: The skill incorporates safety rules in SKILL.md and uses a guardRequest utility in the main execution path to validate user input. It defines a structured narrative workflow and specific output formats to maintain control over the generated content.

gemini-visual-director