{"id":50309782,"url":"https://github.com/josephyaduvanshi/delegate-to-agents","last_synced_at":"2026-05-28T20:01:55.572Z","repository":{"id":359916706,"uuid":"1247997448","full_name":"josephyaduvanshi/delegate-to-agents","owner":"josephyaduvanshi","description":"Delegate coding/research tasks from Claude Code to external agent CLIs (Codex, Gemini, Qwen, OpenClaude, OpenCode) and Claude subagents. Ported from Nous Research hermes-agent.","archived":false,"fork":false,"pushed_at":"2026-05-24T04:01:15.000Z","size":32,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-24T06:12:42.266Z","etag":null,"topics":["agent-orchestration","ai-agents","claude-code","claude-code-plugin","codex","delegation","gemini-cli","qwen"],"latest_commit_sha":null,"homepage":null,"language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/josephyaduvanshi.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":"NOTICE","maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-05-24T03:50:17.000Z","updated_at":"2026-05-24T04:01:18.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/josephyaduvanshi/delegate-to-agents","commit_stats":null,"previous_names":["josephyaduvanshi/delegate-to-agents"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/josephyaduvanshi/delegate-to-agents","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/josephyaduvanshi%2Fdelegate-to-agents","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/josephyaduvanshi%2Fdelegate-to-agents/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/josephyaduvanshi%2Fdelegate-to-agents/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/josephyaduvanshi%2Fdelegate-to-agents/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/josephyaduvanshi","download_url":"https://codeload.github.com/josephyaduvanshi/delegate-to-agents/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/josephyaduvanshi%2Fdelegate-to-agents/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":33624221,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-05-28T02:00:06.440Z","response_time":99,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agent-orchestration","ai-agents","claude-code","claude-code-plugin","codex","delegation","gemini-cli","qwen"],"created_at":"2026-05-28T20:01:52.497Z","updated_at":"2026-05-28T20:01:55.549Z","avatar_url":"https://github.com/josephyaduvanshi.png","language":null,"funding_links":[],"categories":[],"sub_categories":[],"readme":"# delegate-to-agents\n\nA Claude Code plugin for handing work to other AI agents. Use it to send a\ncoding or research task to an external agent CLI (Codex, Gemini, Qwen,\nOpenClaude, OpenCode) or to a Claude subagent, run a few of them in parallel, and\nreview what they hand back before you keep it.\n\nIt's a port of the\n[`autonomous-ai-agents`](https://github.com/NousResearch/hermes-agent/tree/main/skills/autonomous-ai-agents)\nskill from Nous Research's `hermes-agent`. Hermes drives other agents with its\nown `terminal()`, `process()`, and `delegate_task()` tools. Claude Code already\nhas equivalents for all three, so this is mostly translation: the same patterns,\nrewritten against Claude Code's native `Bash` (background mode), `Monitor`, tmux,\nand the `Agent` tool. Nothing runs in the background as a daemon. It's plain\nskill files on top of tools you already have.\n\n## Why bother\n\nClaude Code can already delegate. The `Agent` tool spawns Claude subagents, and\nthere are separate rescue plugins for Codex, Gemini, Qwen, and OpenClaude. What\nwas missing is a single place that ties it together: something that knows when to\nreach for the `Agent` tool versus an external CLI, gives each agent a driver with\nthat agent's actual flags, and keeps the parallel and review patterns in one\nspot. So instead of remembering six different CLIs and which rescue plugin does\nwhat, you say \"delegate this to Codex\" and it routes.\n\n## What's inside\n\n- `delegate-to-agents` — the router. Picks the mechanism and points you at the\n  right driver. Holds the Hermes-to-Claude-Code mechanism mapping so the rest of\n  the skills don't repeat it.\n- `drive-codex`, `drive-gemini`, `drive-qwen`, `drive-openclaude`,\n  `drive-claude-code`, `drive-opencode` — one driver per agent. Each has that\n  CLI's flags, an auth preflight, the one-shot and interactive patterns, and the\n  gotchas worth knowing.\n- `multi-agent-lane` — for when a driven patch needs reviewing before you trust\n  it. Isolate in a git worktree, read the diff, run the tests yourself, then\n  accept, partially accept, or reject. Generalized from Hermes' `kanban-codex-lane`.\n- `cli-delegate` — a Bash-only subagent that drives a CLI to completion and\n  returns a summary, so a long or noisy run never floods your main context.\n- `/delegate` — a one-line entry point: `/delegate codex fix the failing auth test`.\n\n## How it maps to Hermes\n\n| Hermes | Claude Code |\n|--------|-------------|\n| `terminal(command, background=true)` | `Bash(run_in_background=true)` writing to a logfile |\n| `terminal(pty=true)` for a TUI | tmux driven through `Bash` |\n| `process(poll / log / kill)` | `tail` the logfile, or `tmux capture-pane`; kill the shell or session |\n| `process(wait)` | let the background `Bash` exit; the harness re-invokes you |\n| `delegate_task(goal)` | `Agent(subagent_type, prompt)` |\n| `delegate_task(tasks=[...])` | several `Agent` calls in one message |\n\nThe `delegate-to-agents` skill has the full table.\n\n## Install\n\n```\n/plugin marketplace add josephyaduvanshi/delegate-to-agents\n/plugin install delegate-to-agents@delegate-to-agents\n```\n\nOr try it for a single session without installing, by cloning the repo and\npointing Claude Code at the plugin directory:\n\n```\ngit clone https://github.com/josephyaduvanshi/delegate-to-agents\nclaude --plugin-dir delegate-to-agents/plugins/delegate-to-agents\n```\n\n## Use\n\nPlain language works once the plugin is installed:\n\n\u003e delegate the auth refactor to Codex and keep my context clean\n\nSo does the command:\n\n```\n/delegate gemini review src/server.ts for security issues\n/delegate --review codex migrate the tests to pytest\n/delegate --bg qwen add integration tests for the API\n```\n\n`--review` runs the task in an isolated worktree and walks you through the\nreview-and-reconcile checklist before anything is kept. `--bg` runs it in the\nbackground. Ask for several independent tasks at once and they go to separate\nsubagents, one git worktree each, so they don't step on each other.\n\n## A note on the flags\n\nEvery flag in the drivers was checked against the CLI version installed when this\nwas written: Codex 0.133.0, Gemini 0.43.0, Qwen 0.15.6, OpenClaude 0.14.0, Claude\nCode 2.1.150. These tools change quickly, so each driver tells you to re-run\n`\u003ccli\u003e --help` if something looks off. A few that already drifted from the\nupstream Hermes skill: Codex `exec` dropped `--full-auto`/`--yolo` in favor of\n`-s` sandbox modes, Gemini uses `auto_edit` while Qwen uses `auto-edit`, and\nQwen needs `--chat-recording` before a session can be resumed.\n\n## Credits\n\nAdapted from the `autonomous-ai-agents` skill in\n[NousResearch/hermes-agent](https://github.com/NousResearch/hermes-agent) (MIT).\nThe orchestration patterns and the per-agent driver structure are theirs; the\nClaude Code translation, the verified flags, and the extra drivers are this\nplugin's. Licensed under MIT.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjosephyaduvanshi%2Fdelegate-to-agents","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fjosephyaduvanshi%2Fdelegate-to-agents","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjosephyaduvanshi%2Fdelegate-to-agents/lists"}