The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has a wide surface area for data ingestion as it processes user-supplied code, database content, and API responses (e.g., in references/cloud-functions/SKILL.md). However, it provides robust boundary markers and mandatory sanitization protocols such as the 'Sensitive Runtime Data Protection' protocol to prevent accidental command execution or data leakage.
[EXTERNAL_DOWNLOADS]: The skill includes instructions to download official SDKs and assets from trusted sources like Tencent Cloud's official domains (tencentcloudapi.com, cloudbase.net) and well-known repositories on GitHub (github.com/TencentCloud). These are documented neutrally as required resources for TCB development.
[COMMAND_EXECUTION]: The skill provides instructions for using CLI management tools such as the tcb CLI and wechatide for project deployment, debugging, and resource management. These operations are intended for the developer's environment and are accompanied by explicit safety warnings and dry-run procedures.
[DYNAMIC_EXECUTION]: The skill contains deployment scripts and bootstrapping files (e.g., scf_bootstrap, environment build scripts) required for serverless functions and containerized services. These follow standard industry practices for TCB deployments and do not involve executing untrusted dynamic content.
[DATA_EXFILTRATION]: The skill explicitly forbids the wholesale mirroring of request headers or environment variables to prevent the leakage of temporary cloud credentials like x-cloudbase-context. It mandates the use of an allowlist for any debug or echo responses.

cloudbase