The Agent Skills Directory

Unverifiable Dependency (Potential Supply Chain Risk): The documentation for the Python SDK (python/claude-api/README.md and python/claude-api/sdk-upgrade.md) includes instructions to use a library called httpx2 instead of the standard httpx. It claims that httpx2 is an official fork maintained by Pydantic. This is a security consideration as the official Anthropic SDK relies on httpx, and httpx2 is not a standard or verified dependency in this context. Following these instructions could lead to using an unofficial package.
Sensitive Credential Management: The skill provides guidance on using Admin API keys (sk-ant-admin...) for organization management and cost optimization. These credentials grant significant permissions, and their inclusion in documentation serves as a reminder to handle such keys with high security standards, such as using environment variables as recommended in the text.
Data Ingestion Surface: The prompt-audit and migrate subcommands are designed to analyze and modify existing codebases and prompts. This functionality requires the skill to process user-supplied data, which is an inherent part of its purpose as a development utility but represents an ingestion surface for potentially untrusted content.
Use of Trusted Sources: All external documentation and repository links point to official domains (platform.claude.com) or the author's official GitHub organization (github.com/anthropics/*). These are recognized as legitimate and trusted sources for information.

claude-api