The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions in SKILL.md include a 'Required evidence gate' that forces the agent to read and follow instructions from references/cases.md. While currently benign, this structure creates a surface for indirect prompt injection where instructions embedded in the 'Case evidence' could override the agent's primary safety guardrails.
[EXTERNAL_DOWNLOADS]: The skill references multiple external media assets and case studies from goodcase.ai, youmind.com, and x.com (Twitter). These are documented as reference anchors for video generation and do not involve executable code or automated downloads to the local environment, following the intended purpose of the skill.
[DATA_EXPOSURE_&_EXFILTRATION]: The skill contains a safety rule explicitly forbidding the publishing, purchase, or upload of private assets and making external account changes without user approval, which aligns with security best practices for AI agents.

action-continuity-choreography