dify-docs-translation-test
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs a sub-agent to process the content of documentation files (
{SOURCE}and{DRAFT}), which represents an attack surface for indirect prompt injection. If these files contain adversarial instructions, they could influence the agent's review verdict or behavior. - Ingestion points: The content of external English and translated pages is interpolated into the dispatch prompt in SKILL.md.
- Boundary markers: No specific delimiters or wrapping tags are used for the external file content, relying instead on natural language instructions to limit the agent's scope.
- Capability inventory: The sub-agent is designated as 'general-purpose', though the instructions explicitly restrict it from executing other commands.
- Sanitization: The skill does not implement sanitization or filtering of the content retrieved from the documentation files.
Audit Metadata