init

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill performs standard repository initialization by creating a CLAUDE.md file, which is a benign documentation task without malicious intent.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from codebase files such as README.md and .cursorrules. 1. Ingestion points: Repository files (README.md, .cursorrules, .github/copilot-instructions.md). 2. Boundary markers: Absent for input data. 3. Capability inventory: Limited to file creation (SKILL.md). 4. Sanitization: Absent. The vulnerability surface is acceptable given the limited capabilities and specific focus on generating documentation.\n- [DATA_EXFILTRATION]: The skill mentions sensitive configuration paths like ~/.gemini/settings.json but explicitly instructs the agent not to read these files directly, instead directing the user to use secure platform-native import tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 11:30 PM
Security Audit — agent-trust-hub — init