huawei-cloud-swr-image-management

Pass

Audited by Gen Agent Trust Hub on Aug 27, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONPRIVILEGE_ESCALATION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The installation guide in references/cli-installation-guide.md provides commands to download and execute the official hcloud CLI installation script from a Huawei Cloud domain using a curl to bash pipeline.
  • [COMMAND_EXECUTION]: The skill uses hcloud CLI commands to interact with cloud services. It mitigates risk by enforcing a mandatory confirmation workflow for all create, update, and delete operations as described in SKILL.md.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from API outputs such as repository names and tag descriptions in task-repository-management.md and task-tag-management.md. While boundary markers are absent, the risk is lowered by the human-in-the-loop requirement for destructive actions.
  • [DATA_EXFILTRATION]: The skill accesses Docker registry authentication tokens via CreateAuthorizationToken and CreateSecret APIs. It implements security guardrails by explicitly instructing the agent in SKILL.md not to decode these tokens, instead passing them to the user for local terminal execution.
  • [PRIVILEGE_ESCALATION]: The installation instructions in references/cli-installation-guide.md include the use of sudo for moving the hcloud binary to system paths, representing a standard software installation process.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 27, 2026, 08:20 PM
Security Audit — agent-trust-hub — huawei-cloud-swr-image-management