laohan-cheat

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local shell script (bridge.sh) found within the user's home directory (~/.openclaw/workspace-shared/content-ops/) when the OpenClaw environment is detected. This is a functional requirement for its stated integration purpose.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. It ingests untrusted data from user-supplied script files and processes them through an LLM for scoring and auditing (using the mcp__llm-chat__chat tool). There are no explicit boundary markers or 'ignore' instructions implemented to prevent the agent from executing commands hidden within the scripts being analyzed.
  • Ingestion points: Reads files from scripts/ directory and rubric_notes.md.
  • Boundary markers: Absent; untrusted content is interpolated directly into analysis prompts.
  • Capability inventory: Bash execution, file write/edit, and cross-model chat capabilities.
  • Sanitization: No evidence of content sanitization or escaping before prompt interpolation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 07:30 AM
Security Audit — agent-trust-hub — laohan-cheat