The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a utility script scripts/validate_plan.py that executes the vcr command. This is implemented using subprocess.run with an argument list rather than a shell string, which is a secure practice that prevents shell injection vulnerabilities.
[DATA_EXPOSURE]: The validation logic in the skill ensures that output files specified in the render plans must have a .mov extension. This prevents the agent from being manipulated into overwriting critical system files or scripts through the rendering process.
[PROMPT_INJECTION]: The SKILL.md file defines a highly constrained role for the agent with specific response formats and capability boundaries. It includes explicit instructions on how to handle unsupported or impossible requests, reducing the likelihood of the agent being coerced into unintended behaviors.
[SAFE]: The analysis did not detect any hardcoded credentials, unauthorized network exfiltration, obfuscated code, or suspicious persistence mechanisms across the skill files.

vcr-render-planner