The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands and local scripts to automate the research infrastructure.
The script scripts/init-corpus.sh is used to deterministically scaffold the directory structure for the research project.
The orchestration logic involves dispatching multiple sub-processes through the codex CLI and shell utilities like xargs to perform parallel tasks as described in references/codex/wave-dispatch.md.
[REMOTE_CODE_EXECUTION]: The skill is designed to execute autonomous sub-agents with extensive system capabilities.
Documentation in references/codex/codex-exec-contract.md specifies the use of the --dangerously-bypass-approvals-and-sandbox flag for codex exec. This intentional configuration disables the security sandbox and interactive approval gates to allow sub-agents to perform unhindered web research and filesystem writes, which is noted as the sanctioned mode for this skill's high-scale orchestration.
[EXTERNAL_DOWNLOADS]: The skill references and installs external components.
The INSTALL.md file provides commands to download the skill pack using npx from the author's GitHub repository.
The system depends on the external codex-cli (version 0.130.0 or later) for executing research jobs.

run-deep-research