openclaw-desk-pet-macos

Warn

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill requires cloning a repository from a third-party GitHub account (LeoZhaorx/openclaw-desk-pet) which is not a verified or well-known organization.
  • [COMMAND_EXECUTION]: The installation and usage instructions involve executing several shell scripts and command files from the downloaded repository (e.g., ./start-desk-pet.command, ./launch.sh, ./stop-desk-pet.command). This constitutes execution of unverified external code.
  • [INDIRECT_PROMPT_INJECTION]: The skill implements a FileFallbackMonitor that reads and parses local session logs (sessions.json, log.jsonl) to visualize agent states. These files are populated by agent activity which may include untrusted data from external sources.
  • Ingestion points: FileFallbackMonitor.getCurrentState() reads data from sessions/sessions.json and log.jsonl.
  • Boundary markers: None detected in the provided Swift integration patterns.
  • Capability inventory: The skill has the capability to execute shell commands (via repo scripts) and perform network requests to a local or remote gateway.
  • Sanitization: ResultBubbleView performs basic string replacement for control markers like [THINKING] but does not implement comprehensive validation for logic-based injection.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 12, 2026, 06:50 PM
Security Audit — agent-trust-hub — openclaw-desk-pet-macos