story-review

Pass

Audited by Gen Agent Trust Hub on May 13, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by reading untrusted story content and passing it directly to sub-agents for analysis without security boundaries.
  • Ingestion points: Story text is retrieved from local files or through the git diff command in Phase 1.
  • Boundary markers: The prompts for sub-agents (story-architect, character-designer, narrative-writer, consistency-checker) interpolate the story content using the {待审查内容} placeholder without any delimiters or instructions to ignore embedded commands.
  • Capability inventory: The skill has the ability to spawn sub-agents and read local project configuration and rubric files.
  • Sanitization: There is no evidence of content validation, escaping, or filtering of the story text before it is processed by the AI agents.
Audit Metadata
Risk Level
SAFE
Analyzed
May 13, 2026, 06:34 AM