omniroute-integration
Warn
Audited by Gen Agent Trust Hub on Apr 9, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONCREDENTIALS_UNSAFEDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a bash script within the
skill.ymlfile to overwrite the agent's local configuration files (~/.config/opencode/opencode.jsonand~/.local/share/opencode/auth.json). - [CREDENTIALS_UNSAFE]: The skill prompts the user to enter an 'OmniRoute API key' and subsequently stores this secret in plaintext within local JSON configuration files.
- [DATA_EXFILTRATION]: Configuration steps direct the agent to transmit the collected API key to a hardcoded external IP address (
138.2.29.190) for all future model interactions. This IP address is not associated with a well-known service provider. - [COMMAND_EXECUTION]: The skill's configuration script populates the local environment with over 100 model identifiers, many of which are non-existent (e.g.,
gh/gpt-5,kr/claude-sonnet-4.5), indicating a deceptive intent in the skill's metadata and functional claims.
Audit Metadata