The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides detailed instructions for executing the claude CLI as a subprocess using the bash tool. It includes specific configurations to ensure reliable execution in non-interactive environments, such as unsetting specific environment variables (CLAUDECODE, CLAUDE_CODE_ENTRYPOINT) and redirecting stdin from /dev/null to prevent process hangs.
[COMMAND_EXECUTION]: The documentation mandates the use of the --dangerously-skip-permissions flag. This is an official Claude Code CLI flag that enables autonomous operation by bypassing per-tool-call user confirmation prompts, allowing child agents to execute commands and modify files without manual intervention.
[COMMAND_EXECUTION]: The skill describes patterns for fan-out execution, where a parent session spawns multiple concurrent background processes or manages them via generated shell scripts.
[COMMAND_EXECUTION]: It provides instructions for introspecting child sessions by directly reading and parsing Claude Code's internal JSONL session logs located in ~/.claude/projects/ using standard Unix utilities like grep and jq.

nest-claude