The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches latest API documentation, model catalogs, and pricing information from official Anthropic and Claude platforms (platform.claude.com).
[COMMAND_EXECUTION]: Provides instructions and tools for executing shell commands and performing local file modifications through the Agent SDK's built-in tools like Bash, Edit, and Write.
[PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection due to its data ingestion capabilities.
Ingestion points: The skill is designed to read external documentation via WebFetch (URLs listed in shared/live-sources.md) and local project files using Read, Glob, and Grep tools.
Boundary markers: The skill body does not specify explicit delimiters or markers to separate untrusted data from system instructions during tool use.
Capability inventory: Includes subprocess execution (Bash), filesystem writes (Write, Edit), and the ability to spawn subagents (Agent).
Sanitization: File analysis instructions in python/claude-api/tool-use.md correctly recommend the use of os.path.basename() to mitigate path traversal risks when handling output files.
[REMOTE_CODE_EXECUTION]: Implementation examples include the use of npx to dynamically download and execute Model Context Protocol (MCP) servers from official repositories such as Playwright and the Model Context Protocol organization.

claude-api