The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines functions that interpolate user-provided data into Python scripts for Nuke, creating a potential surface for indirect instructions. 1. Ingestion points: The parameters plate_path, output_path, and keyer_type in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: The skill uses nuke.createNode and node['file'].setValue in SKILL.md and has the Bash tool enabled. 4. Sanitization: Absent; input values are not validated before being used in scripts.
[DYNAMIC_EXECUTION]: The function setup_keying_comp utilizes nuke.createNode(keyer_type), allowing for dynamic instantiation of node types based on variable input, which is a common but dynamic execution pattern in this context.

nuke-compositing