The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references official Python and TypeScript SDKs from Anthropic. All installation commands point to standard packages (e.g., anthropic, claude-agent-sdk, @anthropic-ai/sdk) on official registries like PyPI and NPM. References to GitHub repositories point exclusively to the anthropics organization.
[CREDENTIALS_UNSAFE]: The code examples demonstrate secure practices for handling API keys, such as retrieving them from environment variables (os.environ.get('ANTHROPIC_API_KEY')). There are no hardcoded secrets; placeholders like 'your-api-key' are used appropriately for documentation.
[COMMAND_EXECUTION]: The skill documents the Bash and Edit tools within the Claude Agent SDK context. It includes a functional reference for permission_mode settings (e.g., acceptEdits, bypassPermissions), which are standard features of the SDK used to manage tool execution safety. These examples are purely educational.
[DATA_EXFILTRATION]: While the skill describes network-capable tools (WebSearch, WebFetch), they are presented as part of the SDK's standard toolset for building agents. No malicious exfiltration patterns or suspicious external domains are present.

reference-anthropic-sdk