ordercli
Warn
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on the execution of the ordercli binary for all operations, including account login, order tracking, and session management.
- [EXTERNAL_DOWNLOADS]: Instructions specify the installation of the ordercli tool from external sources managed by the author, specifically through Homebrew and Go modules (github.com/steipete/ordercli).
- [CREDENTIALS_UNSAFE]: The skill manages highly sensitive authentication artifacts, including session cookies and browser profile directories (e.g., ~/Library/Application Support/ordercli/browser-profile) used to bypass bot protection mechanisms.
- [PROMPT_INJECTION]: The skill processes untrusted order data from external delivery service APIs which is then interpolated into the agent's conversation context. 1. Ingestion points: Order history and status details retrieved from Foodora and Deliveroo APIs in SKILL.md. 2. Boundary markers: No explicit delimiters or instructions to ignore embedded malicious content are present in the prompt templates. 3. Capability inventory: Shell command execution via the ordercli tool for data retrieval, cart management, and session imports. 4. Sanitization: No methods for filtering, escaping, or validating the external API content are described.
Audit Metadata