clawtributor

Pass

Audited by Gen Agent Trust Hub on Apr 18, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill documents malicious prompt patterns to assist the agent in identifying threats for reporting. These are reference examples, not instructions to override the agent's behavior.
  • [DATA_EXFILTRATION]: Reports are stored locally and submitted manually via a browser-based form only after explicit user approval. Privacy checklists are included to prevent sensitive data from leaving the host.
  • [COMMAND_EXECUTION]: Local file management is performed using standard commands, including chmod 600 to secure report files. These operations are transparent and follow security best practices.
  • [SAFE]: Indirect Prompt Injection Surface Evaluation: 1) Ingestion: Observing suspicious activity (SKILL.md); 2) Boundaries: Explicit user approval checkpoints (SKILL.md); 3) Capabilities: Local file system writes to ~/.openclaw/ (SKILL.md); 4) Sanitization: Mandatory removal of user data and credentials from reports (reporting.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 18, 2026, 03:49 PM