The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and process untrusted storyboard scripts or verbal descriptions from users to populate its internal schema.
Ingestion points: User-provided storyboard scripts and descriptions processed in the '工作流程' section of SKILL.md.
Boundary markers: No delimiters or 'ignore embedded instructions' warnings are present to isolate the processed data from agent instructions.
Capability inventory: The skill performs text transformation and JSON formatting; it does not have shell execution or network access capabilities.
Sanitization: There is no evidence of sanitization or escaping of the user-provided content before interpolation into the final output prompts.
[DATA_EXFILTRATION]: The skill instructions reference access to a specific local file path for retrieval of examples.
Evidence: The SKILL.md file points to ~/.claude/case-library/ as a source for existing case references.
Context: This involves localized filesystem access within the agent's expected working environment; however, it does not target sensitive system credentials or include instructions for remote exfiltration.
[NO_CODE]: The skill package does not contain any executable scripts or binary files, relying entirely on Markdown-based instructions and a static JSON schema for its operation.

video-prompt-schema