The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references official documentation and resources from well-known technology organizations including Anthropic (anthropic.com, code.claude.com) and the Agent Skills open standard (agentskills.io). It also points to the anthropics/skills repository on GitHub for examples. These are trusted sources.- [COMMAND_EXECUTION]: The documentation explains the 'Dynamic Context Injection' feature, which uses the !command syntax to execute shell commands before sending content to the AI. The provided examples demonstrate legitimate use cases using the GitHub CLI (gh) to fetch pull request data for summarization.- [SAFE]: The skill functions strictly as a reference manual and documentation suite. It contains no executable malicious logic or instructions designed to bypass safety filters. It correctly highlights security best practices, such as using the allowed-tools field for capability scoping and disable-model-invocation for preventing unauthorized use of sensitive skills.

claude-skills-overview-2026