short-drama-assets
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes drama scripts (
剧本.md) which are external, untrusted data sources. This creates a potential surface for indirect prompt injection where malicious instructions could be embedded in the script content to influence the agent's behavior during asset extraction. - Ingestion points: Drama script files (
剧本.md) used as input for asset extraction tasks inSKILL.mdandreferences/occurrence-extraction.md. - Boundary markers: The skill uses structured output formats (markdown sections and specific identification anchors like
识别锚点) to isolate and categorize extracted data. - Capability inventory: The agent is authorized to write to local visual setting files (
视觉设定.md) and interact with other drama production tools such as$short-drama. - Sanitization: Skill instructions explicitly direct the agent to focus strictly on visible facts, avoid narrative summaries, and strictly follow project-defined policies for sensitive text, brand names, and URLs as described in
references/prop-and-state.md.
Audit Metadata