natural-writing-ja
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to have an agent revise or write Japanese text based on specific stylistic norms. However, it does not include instructions for the agent to use boundary markers or to ignore embedded commands within the text being revised, which creates a surface for indirect prompt injection attacks. Ingestion points: User-provided Japanese drafts (referenced as draft.md). Boundary markers: Absent. Capability inventory: The skill consists of instructional content and does not define access to tools, network operations, or file system writes in its metadata. Sanitization: Absent.
- [DYNAMIC_EXECUTION]: The skill documentation refers to a Python script (slopscore.py) located at a relative path (../ai-index/scripts/) outside the skill's root directory. While intended for manual draft verification, referencing scripts at external relative paths can be problematic if an agent attempts to execute them autonomously, as the script's integrity is not managed by the skill itself.
Audit Metadata