story-review
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, obfuscated payloads, or unauthorized network operations were detected across the skill's instructions or its 11 supporting scripts. The skill's behavior is entirely consistent with its purpose as a fiction development and review tool.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute several local scripts included in its package (e.g.,
node scripts/language_gate.js,python scripts/tracking_commit.py). These scripts perform automated prose linting, punctuation normalization, and state management. The execution is confined to the local project environment and uses standard runtimes. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied fiction text (
正文), which serves as a potential surface for indirect prompt injection. However, the skill implements a defensive 'Gate' architecture. Before content is processed by the LLM sub-agents, it is scanned by deterministic scripts (language_gate.js,check-style-hygiene.js, andcheck-ai-patterns.js) that identify and block hidden markup, language leaks, and known adversarial patterns.
Audit Metadata