The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from After Effects projects, creating a surface for indirect prompt injection attacks where malicious instructions hidden in text layers or expressions could influence the agent.
Ingestion points: The skill uses ae_text to read text content from layers, ae_find_animation to read animation expressions, and ae_render_frame to visually inspect content for text beats in SKILL.md.
Boundary markers: There are no instructions for the agent to use delimiters or ignore embedded instructions when evaluating extracted text.
Capability inventory: The agent has access to file system operations (ae_save_project, ae_render_video), project modification (ae_item_action, ae_duplicate_comp), and shell command execution (ffprobe).
Sanitization: No sanitization or validation of extracted content is performed before it is processed by the agent.
[COMMAND_EXECUTION]: The skill instructs the agent to use the standard system utility ffprobe to verify the technical metadata of rendered output files.
Evidence: Usage of ffprobe -v error -show_entries stream=width,height,r_frame_rate:format=duration -of json out.mp4 in SKILL.md.

ae-deliver