The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external video data, including transcriptions and visual content. This creates a potential surface for indirect prompt injection if the source media contains malicious instructions designed to influence the agent's QA verdict. However, the skill implements a structured validation process and a rigid JSON report schema which helps mitigate accidental or adversarial influence.
[COMMAND_EXECUTION]: The skill correctly instructs the agent to use ffmpeg and ffprobe for deterministic checks (e.g., blackdetect, freezedetect). These are standard, domain-appropriate tools for video processing and are used with specific, non-malicious parameters.
[SAFE]: No evidence of hardcoded credentials, unauthorized network exfiltration, persistence mechanisms, or obfuscation was detected. The skill follows best practices by requiring a final delivery gate rather than relying on prior stage success reports.

video-quality