The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes several official VolcEngine SDK packages to perform resource diagnostics. These include 'volcenginesdkcore', 'volcenginesdkbilling', 'volcenginesdkecs', 'volcenginesdkvpc', 'volcenginesdkclb', 'volcenginesdkalb', and 'volcenginesdkvke'. These are verified vendor resources provided by the skill author to facilitate service integration.
[COMMAND_EXECUTION]: The diagnostic logic relies on the execution of local CLI tools, specifically 've' (VolcEngine CLI) and 'tosutil' (TOS management tool). The instructions provided across the domain guides (e.g., 'compute-container-network/references/query-cli-catalog.md') explicitly define command packages used to query resource state. These operations are restricted to read-only actions (Describe/List/Get) by default, with any state-changing operations requiring explicit human-in-the-loop confirmation.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to ingest untrusted data from the user and external system logs to perform analysis.
Ingestion points: Untrusted data enters the agent context through user-supplied parameters such as 'RequestId', 'InstanceId', and error message snippets (documented in 'SKILL.md' and various domain 'README.md' files).
Boundary markers: The instructions include boundary markers and instructions to the agent to mask sensitive information (e.g., 'account-permission/references/06-access-key/README.md' requires masking AccessKeyIds).
Capability inventory: The skill possesses capabilities for subprocess execution through 'RunSkillCLI' and 'RunSkillScript', and network operations via the integrated SDKs.
Sanitization: While the skill relies on standard SDK/CLI parameter passing, the markdown instructions prioritize manual verification and explanation of findings to the user, providing a layer of oversight.

volcengine-troubleshooting