The Agent Skills Directory

[SAFE]: The skill contains comprehensive documentation and production-ready examples for creating autonomous agents, focusing on structural best practices and design patterns.
[COMMAND_EXECUTION]: Includes scripts/validate-agent.sh, a local shell script that verifies agent file structure and formatting requirements without performing network requests or accessing sensitive system data.
[PROMPT_INJECTION]: Establishes guidelines for building agents that ingest external codebase information, which presents an indirect prompt injection surface.
Ingestion points: Reference agents use file-system tools (Read, Glob) to pull content from the project directory into the model's context.
Boundary markers: Current templates do not include specific delimiters or 'ignore' instructions for processed data.
Capability inventory: Example agents are configured with access to tools including Write, Bash, Read, and Glob.
Sanitization: No explicit validation or escaping of ingested source code is defined in the provided examples.

agent-development