cloud-engine-canisters
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [SAFE]: The skill serves as a technical reference for developers working with CloudEngine subnets on the Internet Computer. It provides legitimate instructions for handling cycles and cross-subnet calls according to the platform's free cycles cost schedule.
- [INDIRECT_PROMPT_INJECTION]: The skill describes functionality for canisters to make HTTPS outcalls to external APIs (Rule 3, SKILL.md). Ingestion point: HTTP response data. Boundary markers: The IC platform requires a transform function on the calling canister to sanitize or normalize response data. Capability inventory: The skill covers inter-canister messaging and management canister calls. Sanitization: Platform-level transform functions are the standard method for handling external data. This reflects standard architectural patterns of the Internet Computer and not an adversarial injection.
- [COMMAND_EXECUTION]: The skill mentions CLI tools and identity linking for deployment but provides no instructions for executing arbitrary or malicious shell commands.
- [DATA_EXFILTRATION]: HTTPS outcalls are described for legitimate data retrieval. No instructions exist to read sensitive environment variables or local files and transmit them externally.
- [EXTERNAL_DOWNLOADS]: The skill references platform-specific tools like the Motoko compiler (moc) and Rust CDK (ic-cdk) within their official versions, which are standard for the Internet Computer environment and do not constitute risky unverifiable dependencies.
- [PROMPT_INJECTION]: The content consists of technical documentation and code snippets intended for development assistance. There are no attempts to override agent behavior, bypass safety protocols, or manipulate the underlying LLM's instructions.
Audit Metadata