The Agent Skills Directory

[PROMPT_INJECTION]: The skill documents patterns for processing data from external sources, such as "ML Node output", creating a surface for indirect prompt injection.
Ingestion points: The skill ingests untrusted data through node inputs (input_node) and external ML process outputs in SKILL.md.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the patterns.
Capability inventory: The skill requests Bash tool access and utilizes the Nuke Python API for file and process manipulation.
Sanitization: No sanitization or validation of input data is implemented in the provided code snippets.
[COMMAND_EXECUTION]: The skill uses dynamic execution to instantiate Nuke nodes based on string parameters.
Evidence: The create_node_safely function in SKILL.md uses getattr(nuke.nodes, node_type)() to create nodes.
Context: While standard for Nuke automation, this pattern could allow the creation of nodes capable of executing arbitrary code if the node_type argument is influenced by untrusted input.
[COMMAND_EXECUTION]: The skill requests permission for the Bash tool in its frontmatter, which provides a high-privilege execution environment for shell commands alongside its Python automation patterns.

nuke-node-tree-patterns