The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands to manage agent cards.
It uses codex-subagent to launch specialized worker processes.
It uses a pipe sequence awk ... | codex exec ... to manually extract and execute agent prompts from markdown files.
It references a local script scripts/validate-agent.sh to validate the structure and content of agent definitions.
[SAFE]: The skill follows established platform patterns for agent development, emphasizing the use of configuration profiles to set sandbox_mode and approval_policy, which are security best practices for limiting the scope of sub-agent execution.

agent-development