The Agent Skills Directory

[SAFE]: The skill is a well-documented tool for programmatic video editing. It utilizes a local Python package for schema validation and planning, ensuring that all operations are performed on structured data. The implementation demonstrates standard development practices, including automated testing and clear design specifications. No obfuscation, data exfiltration, or unauthorized command execution patterns were found.
[INDIRECT_PROMPT_INJECTION]: The skill processes external inputs such as topics and URLs to generate video scripts and editing plans. This ingestion of external content constitutes a surface for indirect prompt injection, where instructions embedded in the source material could influence the agent's script generation or editing decisions. This risk is inherent to the skill's primary function and is mitigated by the use of structured data models.
Ingestion points: User-supplied topics and URLs (pages or videos) processed via the vpe CLI tool and analyzed by the agent in SKILL.md.
Boundary markers: The skill uses a structured JSON schema (EditPlan) validated by Pydantic to maintain data integrity.
Capability inventory: The agent is authorized to execute the local vpe CLI tool and use agent-browser to interact with a local image generation playground (inemaimg at http://localhost:8000).
Sanitization: Input is classified by type in detect.py, and the resulting plan is validated against specific guardrails in validate.py.

video-plan-editor