seedance-filter
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill's primary purpose is to assist users in rephrasing content that would otherwise be blocked by moderation systems. It provides a "Repair Method" and a table of "Safer Rewrite Patterns" (e.g., replacing conflict with "staged confrontation"). While the skill claims to avoid bypass tactics, the use of euphemisms to avoid "triggering a content filter" is a form of content guideline circumvention.
- [INDIRECT_PROMPT_INJECTION]: The skill has a high exposure to indirect prompt injection as it is specifically designed to ingest and transform untrusted data that has already been flagged as problematic.
- Ingestion points: Processes user prompts that were "blocked, rejected, silently degraded, or likely to trigger a content filter."
- Boundary markers: There are no specific delimiters or "ignore previous instructions" markers provided to encapsulate the untrusted input being analyzed and rewritten.
- Capability inventory: The skill allows the agent to analyze, transform, and output newly generated text based on the potentially malicious input.
- Sanitization: The instructions lack specific sanitization steps or validation rules for the input text beyond the creative repair logic.
Audit Metadata