Создание автономных агентов в Claude Code

★ 8.6 · ai-agents

agent-development is a Claude Code skill that guides the creation of autonomous agents — self-contained subprocesses that handle complex, multi-step tasks without direct user prompting. It covers the complete agent file format: a Markdown document with YAML frontmatter containing four required fields — name (3–50 chars, lowercase letters and hyphens), description (the triggering logic loaded into context at dispatch time), model (inherit, sonnet, opus, or haiku), and color (blue, cyan, green, yellow, magenta, or red). The skill also explains how to restrict an agent's capabilities via the optional tools array using least-privilege sets like ["Read", "Grep", "Glob"] for analysis or ["Read", "Bash", "Grep"] for testing. System prompt design is covered in depth — second-person voice, step-by-step process, output format, and edge-case handling. Aimed at Claude Code plugin developers who need to separate autonomous background agents from user-initiated commands.