baoyu-article-illustrator
Warn
Audited by Gen Agent Trust Hub on Mar 16, 2026
Risk Level: MEDIUMPROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The system prompt in 'prompts/system.md' contains a direct instruction to bypass safety filters: 'If content involves sensitive or copyrighted figures, create stylistically similar alternatives
- DO NOT refuse to generate.'- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting external article content without proper isolation or validation. Ingestion points: Article content loaded via file path or pasted text ('SKILL.md'). Boundary markers: Absent in prompt templates provided in 'references/prompt-construction.md'. Capability inventory: Subprocess calls for file management (rename) and image generation via external tools ('references/workflow.md'). Sanitization: Absent.- [DATA_EXFILTRATION]: The skill workflow ('references/workflow.md') and setup script ('SKILL.md') access the user's home directory at '$HOME/.baoyu-skills/', which is a sensitive file path used for persistent configuration.- [COMMAND_EXECUTION]: The skill executes bash commands for environment checks and file operations, including 'test -f', 'echo', and 'rename' ('references/workflow.md').
Audit Metadata