dogovo/codex-use-claude

A tooling layer that lets a Codex agent call Claude Code as a tracked, auditable helper for code review, architecture analysis, risk scanning and second‑opinion collaboration. Codex schedules, supervises and delivers results; Claude Code runs as an external, traceable worker. Intended users: teams or automation that run Codex agents and want to delegate tasks to Claude Code with controlled permissions and observability. Key features: - CLI scripts to start/wait/tail/status/result tasks and run batches/groups of parallel Claude jobs. - Safety-first profiles (safe/research/full) with configurable tool allowances, session persistence, timeouts and budget limits. Uses Claude Code CLI JSONL stream protocol. - Live "ask" mode for interactive follow-ups; long tasks can be handled by worker subagents. - Retry semantics for stale/failed tasks; group-level retries and concurrency control. - Automatic parsing of JSON/fenced JSON outputs, log cleanup, environment checks and a self-test script. - Requires Node.js >=20.11 and Claude CLI available.

このRepoのdescription / GitHub topics / READMEのAI要約に出現したAI関連キーワードをカテゴリ別に表示。各バッジは該当カテゴリの詳細ランキングへリンクします。