The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized exfiltration attempts were detected. The skill is purely educational and architectural, providing legitimate guidance for selecting and using Sui SDKs.
[COMMAND_EXECUTION]: The skill contains informational command-line examples for installing and upgrading standard packages (e.g., npm install @mysten/sui, pip install pysui). These are standard developer operations and are not executed autonomously by the agent.
[PROMPT_INJECTION]: The skill describes an architectural pattern where the agent reads local documentation from node_modules/@mysten/*/docs/ to provide version-accurate answers. This constitutes an indirect prompt injection surface.
Ingestion points: node_modules/@mysten/*/docs/llms-index.md (documented in llm-docs.md).
Boundary markers: Absent.
Capability inventory: The skill relies on the agent's file-reading and tool-execution capabilities to provide code assistance.
Sanitization: Absent.
Note: This is an intended design pattern to improve AI accuracy by using project-specific dependencies as a source of truth.

sui-sdks