huawei-cloud-waf-aad-rule-management

Pass

Audited by Gen Agent Trust Hub on Sep 24, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Huawei Cloud API responses (such as rule names and domain configurations) which is then used for diagnostics and decision-making. While this data could be manipulated by an attacker who has access to the cloud environment, the skill effectively mitigates this risk through a rigorous confirmation workflow and strict parameter validation.\n
  • Ingestion points: The agent reads resource data via hcloud query commands defined in SKILL.md and references/verification-method.md.\n
  • Boundary markers: The workflow enforces R-levels (R1/R2) requiring explicit user confirmation for any creation or deletion of rules, ensuring a human-in-the-loop.\n
  • Capability inventory: The skill has the ability to modify WAF and AAD configurations using the hcloud CLI tool.\n
  • Sanitization: Parameters are validated against strict whitelists and format checks (e.g., IP addresses, UUIDs, and specific enumerated values) before being passed to shell commands.\n- [EXTERNAL_DOWNLOADS]: The skill documentation references external binaries, specifically the hcloud CLI and a telemetry tool named skill-quality-cli, and provides structured installation guides.\n
  • The scripts/ensure_cli.sh script is designed to check for local existence and does not download code at runtime, preventing silent remote code execution.\n
  • The installation instructions in references/cli-installation-guide.md point to official vendor domains (myhuaweicloud.com) and mandate SHA256 checksum verification before the user executes any installers.\n- [DATA_EXFILTRATION]: The skill is configured to automatically report quality telemetry via the skill-quality-cli wrapper tool.\n
  • This reporting is a documented feature and uses the vendor's official cloud infrastructure for data collection.\n
  • The telemetry is transparent and can be disabled by the user by setting the SKILL_QUALITY_REPORT=0 environment variable.\n- [CREDENTIALS_UNSAFE]: The skill handles authentication via standard environment variables or local user-managed profiles.\n
  • It explicitly instructs users not to hardcode credentials and forbids the skill or its scripts from writing credentials to disk (e.g., using hcloud configure set).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 24, 2026, 06:09 PM