The Agent Skills Directory

[SAFE]: The skill provides a structured framework and best practices for developing autonomous agents within a local environment, following Anthropic's documentation.
[COMMAND_EXECUTION]: The skill uses the Bash tool but restricts its capability to the mkdir command via allowed-tools, which is necessary for creating the agents/ directory structure and prevents arbitrary command execution.
[DATA_EXFILTRATION]: No network-enabled tools are requested, and the instructions do not contain patterns for accessing sensitive files or transmitting data to external servers.
[PROMPT_INJECTION]: The content focuses on instructional guidance for agent behavior and does not include attempts to bypass safety filters, override system instructions, or extract underlying prompts.
[REMOTE_CODE_EXECUTION]: The skill does not perform remote downloads or execute code from external sources. It references a local validation script (scripts/validate-agent.sh) as part of a standard development workflow.

create-agent