The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines functions to process user-provided messages through AI models, which is an inherent attack surface for indirect prompt injection. 1. Ingestion points: The messages array in generateText and streamText calls (SKILL.md). 2. Boundary markers: No explicit input delimiters or instructions to ignore embedded commands are present in the provided code samples. 3. Capability inventory: The skill can invoke AI models and modify cloud environment settings via callCloudApi. 4. Sanitization: The examples do not demonstrate sanitization or filtering of external content before interpolation into prompts.
[COMMAND_EXECUTION]: The skill requires the agent to perform administrative tasks using the callCloudApi tool to ensure the target environment is ready. Evidence: Use of UpdateAIModel to enable models and CreateAIModel to register custom model groups (SKILL.md). These are legitimate configuration actions within the Tencent CloudBase platform.

ai-model-wechat