The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation provides commands for users to download a ZIP archive from the author's GitHub repository. This archive is used to host a gallery of model architecture images for local inspection.
Evidence: SKILL.md contains instructions to use curl and unzip on a release asset from the BBuf/AI-Infra-Auto-Driven-SKILLS repository.
[OBFUSCATION]: Several URLs in references/diagram-index.json use standard URL percent-encoding to support Chinese characters in file paths.
Evidence: Encoded segments like %E6%B7%B7%E5%85%83... in references/diagram-index.json decode to valid descriptive text related to the model names.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied model names and interpolates them into a shell command to run a resolution script.
Ingestion points: User-provided <model name> in the SKILL.md workflow section.
Boundary markers: None present in the shell command template, though the argument is double-quoted to prevent simple breakage.
Capability inventory: Execution of python3 skills/model-architecture-diagram/scripts/model_architecture_diagram.py via shell.
Sanitization: The resolution script implements a normalize function using regex re.sub(r"[^a-z0-9.\-\u4e00-\u9fff]+", " ", text) which effectively strips potentially malicious shell metacharacters and control tokens before processing the input.

model-architecture-diagram