children-stories

Pass

Audited by Gen Agent Trust Hub on Apr 16, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references a Google NotebookLM resource (ID: 3a99bf34-ed1d-4abf-a932-32789965ac7b) to retrieve specialized craft techniques and storytelling research. This involves querying a well-known cloud service for domain-specific knowledge retrieval.
  • [DATA_EXFILTRATION]: The skill manages several local markdown files within the project directory, including story-blueprint.md, voice-sample.md, character-bible.md, and world-bible.md. These operations are limited to project-related data and are essential for maintaining state across different writing sessions.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing external user-provided data:
  • Ingestion points: It reads content from voice-sample.md to calibrate the writing voice and processes iterative feedback during the 'test' mode (SKILL.md, Mode 3).
  • Boundary markers: No specific delimiters or instructions to ignore embedded commands are included for the ingested files.
  • Capability inventory: The skill possesses file-writing capabilities to generate and update story files in the stories/ directory.
  • Sanitization: The skill does not explicitly specify sanitization routines for the input text samples, though its output is restricted to children's literature formats.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 16, 2026, 11:34 AM