openclaw-desk-pet-macos
Warn
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill facilitates the cloning of a repository from an external, unverified GitHub source (https://github.com/LeoZhaorx/openclaw-desk-pet.git).
- [COMMAND_EXECUTION]: The instructions involve executing scripts contained within the repository, such as start-desk-pet.command and launch.sh.
- [REMOTE_CODE_EXECUTION]: By downloading and immediately running scripts from an external repository not on a trusted list, the skill establishes a pattern of remote code execution.
- [PROMPT_INJECTION]: The skill processes local application logs, introducing a risk of indirect prompt injection.
- Ingestion points: Reads files from ~/.openclaw/sessions/ to monitor agent activity (SKILL.md).
- Boundary markers: Lacks delimiters or instructions to ignore embedded commands when interpreting session log data.
- Capability inventory: Capable of performing network requests to a local gateway and executing shell scripts.
- Sanitization: Does not sanitize or validate log content before using it to influence agent state logic.
Audit Metadata