The Agent Skills Directory

[PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection during its image analysis phase.
Ingestion points: User-provided reference images are decoded into editable JSON specifications in 'PHASE A: ANALYZE' (SKILL.md).
Capability inventory: The skill has 'read+write' capabilities via 'advibly_generate_image', 'advibly_generate_video', and 'advibly_social_create_post' (SKILL.md), allowing interpreted data from images to flow into downstream generation and publishing tasks.
Boundary markers: There are no explicit instructions for the agent to ignore or delimit embedded text/instructions found within visual references.
Sanitization: The workflow lacks validation or sanitization steps for content extracted from user assets before it is used to build prompts.
[SAFE]: All core functionality is mediated through vendor-specific tools ('advibly_*') which are part of the author's ecosystem. This includes image/video generation, asset management, and social posting.
[SAFE]: No hardcoded credentials, persistence mechanisms, or unauthorized privilege escalation patterns were detected in the skill instructions or scripts.
[SAFE]: The skill uses clear, non-obfuscated instructions and relies on informative reference documentation ('references/collage-dimensions.md') to calibrate output quality.

collage-motion