The Agent Skills Directory

[SAFE]: The skill provides documentation and integration guides for the inferencesh Python SDK.
[EXTERNAL_DOWNLOADS]: Instructions for installing the inferencesh package via pip and related CLI tools via npx are standard for this SDK.
[COMMAND_EXECUTION]: Provides examples of tool handlers, including a calculator using eval() and internal code execution capabilities, which are documented as features for developers.
[CREDENTIALS_UNSAFE]: Encourages using environment variables for API keys, which is a secure alternative to hardcoding secrets.
[INDIRECT_PROMPT_INJECTION]: Documents patterns for building agents that ingest user input, specifically highlighting safety mitigations.
Ingestion points: agent.send_message and client.run inputs.
Boundary markers: None mentioned in basic examples.
Capability inventory: SDK provides tool execution, webhooks, and code execution.
Sanitization: Recommends human-in-the-loop approval workflows for sensitive operations.

python-sdk