huawei-cloud-waf-policy-manage
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill uses highly directive language in sections such as 'Initial Interaction (Fixed Procedure)' and 'Execution Discipline (MANDATORY)' to strictly enforce a specific output format and command sequence, which attempts to override the agent's natural instruction processing.
- [INDIRECT_PROMPT_INJECTION]: The skill processes JSON files supplied by the user to automate WAF configuration.
- Ingestion points: The agent reads a JSON file from a user-specified path.
- Boundary markers: No explicit markers or 'ignore' instructions are used for the processed data fields.
- Capability inventory: The skill uses bash commands (
hcloud) and file-writing capabilities to create and modify cloud resources. - Sanitization: While basic validation for policy names is mentioned (regex for length and characters), many other fields from the JSON (rule conditions, tags, metadata) are interpolated into shell commands without explicit sanitization.
- [DYNAMIC_EXECUTION]: The skill implements a workflow that dynamically generates temporary JSON files based on the contents of the user-provided policy file. These generated files are then passed to shell commands using the
--cli-jsonInputparameter to handle CLI argument conflicts. - [EXTERNAL_DOWNLOADS]: The documentation includes reference links to download the official Huawei Cloud CLI (
hcloud) from the vendor's authenticated object storage service (OBS).
Audit Metadata