The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides scripts and patterns for batch processing files that ingest untrusted data (filenames and directory paths) from the execution environment, which could be exploited to influence agent behavior if processed in malicious directories.
Ingestion points: batch_export.py and the bash script in SKILL.md ingest file paths via CLI arguments (argv) and shell globbing.
Boundary markers: Absent; the instructions do not include delimiters or warnings to ignore embedded content within filenames or metadata.
Capability inventory: The skill possesses file system read/write capabilities via the bpy (Blender Python API) and os modules, specifically creating directories and writing .glb files.
Sanitization: The bash snippet in SKILL.md utilizes variable quoting ("$blend_file") to prevent simple command injection, but lacks broader validation of the input source integrity.
[DYNAMIC_EXECUTION]: The skill utilizes dynamic script generation to facilitate Blender automation from the command line.
Evidence: The automation example in SKILL.md (section 7) uses a bash here-doc to dynamically generate Python code and pipe it into the Blender executable (`blender --python
<<EOF`). While this is a standard developer pattern for the task, it constitutes runtime code generation.

blender-web-pipeline