The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides a framework for building AI agents that ingest and process data from users, external tools, and webhooks. This architecture creates an attack surface for indirect prompt injection. Ingestion points include agent.sendMessage input, appTool results, and webhookTool responses in SKILL.md, references/agent-patterns.md, and references/tool-builder.md. The SDK relies on developer-defined system_prompt configurations without enforcing specific delimiters for untrusted content. Capability inventory includes application execution, file management, and built-in code execution as documented in SKILL.md and references/agent-patterns.md. Sanitization is left to the implementing developer.\n- [DYNAMIC_EXECUTION]: Documentation and reference files include patterns that enable or demonstrate dynamic code execution, such as the Code Execution Pattern in references/agent-patterns.md and a code snippet using eval() to process mathematical expressions from tool calls in references/tool-builder.md.\n- [COMMAND_EXECUTION]: The skill includes instructions to install additional platform components using npx skills add belt-sh/cli in SKILL.md, which involves downloading and executing remote code via the CLI utility.

javascript-sdk