cloud-api-operations

Pass

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is primarily designed to facilitate the execution of Tencent Cloud control-plane operations using the callCloudApi tool. It provides detailed workflows for actions such as ModifyEnv, DestroyEnv, and AttachRolePolicy, which are high-privilege operations intended for cloud resource management.
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to fetch SDK source code and parameter schemas from official Tencent Cloud repositories. These sources include GitHub (raw.githubusercontent.com/TencentCloud/), Gitee (gitee.com/TencentCloud/), and Tencent's CNB platform (cnb.cool/tencent/cloud/api/sdk/). These are recognized vendor-owned domains used for legitimate technical reference.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes metadata and logs from the cloud environment, such as audit events from cloudaudit and environment information from tcb/DescribeEnvs. This ingestion of untrusted external data (resource names, usernames, or audit logs) constitutes an indirect prompt injection surface.
  • Ingestion points: Operations such as cloudaudit/LookUpEvents, DescribeEnvInfo, and DescribeDBInstanceAttribute (found in references/recipes/cloud-audit-operation-trace.md and references/recipes/pg-instance-spec.md).
  • Boundary markers: The skill defines the "API first" principle in SKILL.md, instructing the agent to resolve names from documented sources (API index) to maintain strict operational boundaries.
  • Capability inventory: The skill has extensive capabilities to modify cloud infrastructure, manage IAM policies, and perform network requests via callCloudApi across various Tencent Cloud services (CAM, TCB, SCF, etc.).
  • Sanitization: No explicit sanitization logic for ingested data is provided in the instructional content, relying on the underlying agent's standard handling of tool outputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 23, 2026, 06:19 AM