The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's full-pipeline.mjs example utilizes execSync from the child_process module to orchestrate the production workflow. This involves executing shell commands with variable interpolation (e.g., using ${PROJECT} derived from command-line arguments). This pattern is susceptible to command injection if the project path or other variables contain shell metacharacters.
[EXTERNAL_DOWNLOADS]: The installation process includes a setup-indextts2-mlx.sh script which is responsible for downloading IndexTTS-2 machine learning models to the user's local directory (~/.local/share/paper-collage-ad/). This involves fetching external binary assets that are required for the skill's voice cloning functionality.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it processes untrusted data from multiple JSON manifests—such as storyboard.json, voice.indextts2.json, and animation.json—and incorporates this data into prompts or uses it to drive automated scripts. For instance, the Brand Color Extraction example shows how extracted color values and product names are interpolated directly into prompts for keyframe generation.
Ingestion points: manifests/storyboard.json, manifests/voice.indextts2.json, manifests/animation.json, manifests/final.json.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the processing logic shown.
Capability inventory: Subprocess execution (execSync), file system access (read/write), and media processing via FFmpeg.
Sanitization: There is no evidence of validation or sanitization for the content within the JSON manifests before they are used in commands or prompts.

paper-collage-ad-production