agtx One-Shot Project Runner
fynnfluegge/agtx
Runs a whole project unattended on an agtx kanban board, decomposing the goal, starting tasks, unblocking workers and merging each result.
Starts a background runner that works through ready batches of tasks from Plans.md one cycle at a time via Breezing, with status and stop controls.
SKILL.md written in Japanese; this summary is our English description.
Install Harness Loop
$ npx skills add Chachamaru127/claude-code-harness --skill harness-loop -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install Chachamaru127/claude-code-harness harness-loop --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/Chachamaru127/claude-code-harness.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills-codex/harness-loop .claude/skills/harness-loop && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "harness-loop" agent skill from https://github.com/Chachamaru127/claude-code-harness/tree/main/skills-codex/harness-loop into .claude/skills/harness-loop/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "harness-loop", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/Chachamaru127/claude-code-harness/tree/main/skills-codex/harness-loopType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add Chachamaru127/claude-code-harness --skill harness-loop -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install Chachamaru127/claude-code-harness harness-loop --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Chachamaru127/claude-code-harness.git skills-src && mkdir -p .agents/skills && cp -r skills-src/skills-codex/harness-loop .agents/skills/harness-loop && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "harness-loop" agent skill from https://github.com/Chachamaru127/claude-code-harness/tree/main/skills-codex/harness-loop into .agents/skills/harness-loop/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "harness-loop", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add Chachamaru127/claude-code-harness --skill harness-loop -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install Chachamaru127/claude-code-harness harness-loop --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Chachamaru127/claude-code-harness.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/skills-codex/harness-loop .cursor/skills/harness-loop && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "harness-loop" agent skill from https://github.com/Chachamaru127/claude-code-harness/tree/main/skills-codex/harness-loop into .cursor/skills/harness-loop/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "harness-loop", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/Chachamaru127/claude-code-harness.git --path skills-codex/harness-loop--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add Chachamaru127/claude-code-harness --skill harness-loop -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install Chachamaru127/claude-code-harness harness-loop --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Chachamaru127/claude-code-harness.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/skills-codex/harness-loop .gemini/skills/harness-loop && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "harness-loop" agent skill from https://github.com/Chachamaru127/claude-code-harness/tree/main/skills-codex/harness-loop into .gemini/skills/harness-loop/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "harness-loop", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install Chachamaru127/claude-code-harness harness-loopInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add Chachamaru127/claude-code-harness --skill harness-loop -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/Chachamaru127/claude-code-harness.git skills-src && mkdir -p .github/skills && cp -r skills-src/skills-codex/harness-loop .github/skills/harness-loop && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "harness-loop" agent skill from https://github.com/Chachamaru127/claude-code-harness/tree/main/skills-codex/harness-loop into .github/skills/harness-loop/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "harness-loop", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add Chachamaru127/claude-code-harness --skill harness-loop -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install Chachamaru127/claude-code-harness harness-loop --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/Chachamaru127/claude-code-harness.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/skills-codex/harness-loop .opencode/skills/harness-loop && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "harness-loop" agent skill from https://github.com/Chachamaru127/claude-code-harness/tree/main/skills-codex/harness-loop into .opencode/skills/harness-loop/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "harness-loop", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
Facts
- Skill name
harness-loop- GitHub stars
- 3.2k
- Token cost
- ~2.2k tokens
- SKILL.md length
- 660 words
- Files
- 1
- Skills in repo
- 25
- Repo updated
- First seen
- Licence
- MIT
At a glance
Starts a background runner that works through ready batches of tasks from Plans.md one cycle at a time via Breezing, with status and stop controls.
- Works in 9 steps: project root の .claude/state/codex-loop/… → 受け取った selection を Plans.md から正規化する → Plans.md から Depends が満たされた cc:TODO /… → …
- Running many independent plan tasks unattended over a long session
- SKILL.md covers ひとことで, たとえると, Quick Reference and 実行コマンド, plus 9 more sections
- Calls codex and bash
What it does
The skill launches a real background process rather than a simulated loop. Each cycle picks one ready batch, meaning tasks marked cc:TODO or cc:WIP whose dependencies are met, hands it to Breezing, checks the result and moves to the next batch. Run state is written under .claude/state/codex-loop/ in the project root.
You start it with $harness-loop all or limit it to a range of task IDs from Plans.md, and tune it with options for maximum cycles, maximum workers per cycle, a named plan, pacing and an executor that falls back to one task per local worker. harness codex-loop status shows progress and stop asks it to halt. Before starting, the agent checks the plan's goal, definition of done and approvals, and it does not treat a successful launch as proof the work is finished. The SKILL.md is written in Japanese.
When your agent uses it
- Running many independent plan tasks unattended over a long session
- Limiting an autonomous run to a range of Plans.md task IDs
- Checking the progress of a running background loop
- Stopping a loop that is partway through its batches
Example prompts
- “$harness-loop all --max-cycles 3”
- “Run the loop on tasks 41.1-41.4 only, with at most four workers.”
- “Show me the status of the running harness loop.”
- “Stop the background loop once the current job finishes.”
Requirements
- The harness command-line tool
- A Plans.md file listing tasks and their dependencies
- Pre-approved tools (allowed-tools): Read, Bash
Workflow steps
9 steps, taken from the first numbered list in SKILL.md.
- project root の .claude/state/codex-loop/ に Harness loop の実行状態を書き出す
- 受け取った selection を Plans.md から正規化する
- Plans.md から Depends が満たされた cc:TODO / cc:WIP を集め、ready batch を作る
- max-workers で ready batch の同時実行数を制限する
- 既定では Breezing executor が ready batch を Lead / Worker / Reviewer 分離で実行する
- executor task の時だけ、互換用 local worker が 1 task per cycle で codex exec を起動する（CODEX_LOOP_TASK_DRIVER=companion の時だけ bash…
- 高リスク task / 2 回目失敗 / plateau 直前では advisor consult を挟む
- ready batch 完了後に review / checkpoint / plateau 判定を行う
- まだ対象タスクが残っていれば、待機後に次サイクルへ進む
What it can do on your machine
Read from SKILL.md and the folder at commit 2b2b748. It shows what the files ask for, not the result of running them.
Tool permissions
Pre-approves these tools, so the agent can use them without asking each time:
ReadBash
From allowed-tools in the SKILL.md frontmatter.
Runs code
Shell commands in SKILL.md call:
codexbash
From the folder's file list and the shell code blocks in SKILL.md.
Network
No URLs in SKILL.md.
From URLs in SKILL.md, links to its own repository left out.
Credentials
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Context cost
Harness Loop loads about 2.2k tokens when it runs. Until then it costs about 77 tokens; SKILL.md has 660 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
Safety
Auto-check: notes
The automated check noted patterns worth knowing about, such as sudo or a known installer.
- NotePre-approves every shell command (allowed-tools: Bash)SKILL.md
allowed-tools: Read, Bash
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
SKILL.md
The full file from Chachamaru127/claude-code-harness at commit 2b2b748, republished under its MIT licence (© Chachamaru127). 660 words, ~2,168 tokens.
Download SKILL.mdSave it as
.claude/skills/harness-loop/SKILL.md (or your agent's skills folder).- name
- harness-loop
- description
- HAR: Codex-native long-running loop runner. Uses a real background runner that executes one ready batch per cycle through Breezing by default, with status/stop controls. Trigger: long-running, loop, autonomous, background, Codex. Do NOT load for: one-shot implementation, normal review, release.
- allowed-tools
- Read, Bash
- description-en
- HAR: Codex-native long-running loop runner. Uses a real background runner that executes one ready batch per cycle through Breezing by default, with…
- description-ja
- HAR: Codex 専用の長時間ループ実行。実際のバックグラウンドランナーが ready batch を Breezing で進め、status / stop で監視できる。長時間、loop、ループ、autonomous、background、Codex で起動。
- kind
- workflow
- purpose
- Run long-lived Codex ready-batch execution loops
- trigger
- long-running, loop, autonomous, background, Codex
- shape
- delegate
- role
- orchestrator
- base
- harness-work
- pair
- harness-sync
- owner
- harness-core
Harness Loop
Codex 版の harness-loop は、説明だけの擬似ループではなく、
実際にバックグラウンドで回るランナーを起動する。
ひとことで
$harness-loop は、1 回だけの実装依頼ではなく、
「今すぐ実行できる未完了タスクのまとまりを、Breezing で自動実行し続ける当番」を起動する入口。
ここでいう ready batch は、Depends が満たされていて、今すぐ並列実行できる cc:TODO / cc:WIP のまとまり。
1 cycle は 1 task ではなく、原則として 1 ready batch を処理する。
たとえると
人が横でずっと見張る代わりに、 「同時に進められる作業をまとめて見つける → Breezing に任せる → 結果を確認する → 次のまとまりへ進む」 を繰り返す監督係を、裏で常駐させるイメージ。
Quick Reference
| 入力 | 動作 |
|---|---|
$harness-loop all | 未完了タスク全体を長時間ループで開始 |
$harness-loop 41.1-41.4 | 範囲を絞って開始 |
$harness-loop JLB3R-02..JLB3R-08 | Plans.md の task ID 順で範囲を絞って開始 |
$harness-loop all --max-cycles 3 | 最大 3 サイクルで停止 |
$harness-loop all --max-workers 4 | 1 cycle の ready batch を最大 4 worker までに制限 |
$harness-loop all --max-workers max | ready batch 内で実行可能なタスク数を上限として並列化 |
$harness-loop all --plan roadmap | named Plans の roadmap を対象にループ実行 |
$harness-loop all --executor task | 旧来の 1 task per cycle local worker 実行へ逃がす |
$harness-loop all --pacing night | サイクル間の待機を長めにする |
$harness-loop status | 現在の実行状況を確認 |
$harness-loop stop | 進行中ジョブを止めてループ停止要求を出す |
実行コマンド
開始
harness codex-loop start all範囲指定:
harness codex-loop start 41.1-41.4 --max-cycles 5 --pacing worker
harness codex-loop start JLB3R-02..JLB3R-08 --max-cycles 5 --pacing worker
harness codex-loop start all --max-workers max --pacing worker
harness codex-loop start all --plan roadmap --max-cycles 5
harness codex-loop start all --executor task --max-cycles 5START..END は、Plans.md に並んでいる task ID をそのまま使う範囲指定。
英字やハイフンを含む task ID は .. を優先する。
41.1-41.4 のような従来の数値レンジも引き続き使える。
--max-workers は、Breezing が 1 cycle で同時に動かす worker 数の上限。
max は、選択範囲内で Depends が満たされた ready task の数をそのまま上限にする。
--executor task は、Breezing ではなく local worker に 1 task だけ渡す互換用の逃げ道。
問題切り分けや、並列実行したくない危険な作業で使う。
複数 Plans.md がある repo では、長時間 run の起動時に --plan NAME を明示する。
runner は開始時に解決した Plans file を cycle 間で保持するため、途中で active plan を切り替えない。
起動前に、選択済み plan の目的、担当範囲、DoD、spec、検証証拠、原依頼と承認の参照を確認する。状態確認だけの依頼で start を呼ばない。
再開時は最新の run state と選択済み plan を読み、決定と理由、失敗履歴、残る判断、直前の advisor response を回収する。助言で元の契約や承認境界を置き換えない。
不足は読み取りで補い、軽微な仮定を明示する。承認済み可逆作業は再確認せず進め、保護操作や STOP は既定の停止条件を守る。
起動成功は作業完了の証拠ではない。status、job の実結果、DoD と review の証拠を照合して報告する。必須チェック後の追加検証は新しい変更、失敗、未解決の懸念がある場合に限る。
状態確認
harness codex-loop status
harness codex-loop status --json停止
harness codex-loop stopどう動くか
- project root の
.claude/state/codex-loop/に Harness loop の実行状態を書き出す - 受け取った selection を Plans.md から正規化する
- Plans.md から Depends が満たされた
cc:TODO/cc:WIPを集め、ready batch を作る --max-workersで ready batch の同時実行数を制限する- 既定では Breezing executor が ready batch を Lead / Worker / Reviewer 分離で実行する
--executor taskの時だけ、互換用 local worker が 1 task per cycle でcodex execを起動する（CODEX_LOOP_TASK_DRIVER=companionの時だけbash "${HARNESS_PLUGIN_ROOT}/scripts/codex-companion.sh" task --background --write ...を使う）- 高リスク task / 2 回目失敗 / plateau 直前では advisor consult を挟む
- ready batch 完了後に review / checkpoint / plateau 判定を行う
- まだ対象タスクが残っていれば、待機後に次サイクルへ進む
Realtime Handoff / Silence Policy
Codex 0.123.0 以降の background agent は realtime handoff で transcript delta を受け取れる。
この delta は「状況把握用の追記」であり、毎回ユーザーへ返答する合図ではない。
ひとことで: background agent は、必要な時だけ報告し、何も判断が変わらない時は明示的に沈黙する。
たとえると、見張り役が廊下でずっと実況するのではなく、異常・完了・判断待ちだけを知らせる形。
報告してよいタイミング:
- loop 開始、停止、
already running、stop受理など、ユーザー操作に関わる lifecycle 境界 - 1 ready batch cycle の最終結果、commit、
RESULT: APPROVED/RESULT: BLOCKED - Breezing Lead が task 完了を progress feed としてまとめて出す時
- task が blocked、validation failure、review
REQUEST_CHANGES、plateau、advisorSTOPで止まる時 - user が
statusを実行した時、または明示的に途中状況を聞いた時 - advisor / reviewer drift、contract readiness failure など、放置すると品質判定がずれる時
沈黙するタイミング:
- transcript delta を受け取っただけで、task / review / advisor の状態が変わっていない時
runner.log/jobs/*.logに既に残る細かな stdout だけが増えた時pacing待機中で、次 cycle まで新しい判断材料がない時
途中報告の頻度:
- default は「1 ready batch cycle につき最終報告 1 回」。
- Breezing の task-level progress feed は、batch 内の完了数が動いた時だけ出す。
- 長い cycle でも、material state change がない限り heartbeat は出さない。
- 詳細な流れは
harness codex-loop status --jsonと project root の.claude/state/codex-loop/runner.logに寄せ、会話側には要点だけ出す。
Advisor / Reviewer drift との関係:
- silence policy は drift 検知を弱めるためのものではない。
advisor-request.v1に response がない、review-result.v1が返らない、contract が未承認などの異常は必ず state / log に残し、必要ならユーザーへ報告する。- Advisor は
PLAN/CORRECTION/STOPの相談役、Reviewer は最終品質判定役のまま分離する。
Show full SKILL.md (233 more words)Show less
pacing
| 値 | 用途 | 待機秒数 |
|---|---|---|
worker | 通常の開発ループ | 270 |
ci | 短めに確認したい時 | 270 |
plateau | 行き詰まり気味の再試行 | 1200 |
night | 長めの放置実行 | 3600 |
State Path Policy
Codex 版 harness-loop は、Codex native の会話・実行キャッシュと、Harness が共有する project state を分けて扱う。
- Harness 共通 state: project root の
.claude/state/配下に置く。Claude 側の advisor / review / checkpoint と共有するため、harness codex-loop statusもここを読む。 - Codex loop runner state: project root の
.claude/state/codex-loop/配下に置く。これは「Codex 全体の正本」ではなく、Harness loop runner の job / cycle / log 用 state。 - Codex native state:
${CODEX_HOME:-~/.codex}配下に残る Codex 自身の thread / transcript / cache。Harness loop の task status、advisor history、review result の正本にはしない。 - 禁止:
.Codex/や~/.Codexを正本 path として案内しない。大文字Codexディレクトリは historical drift と見なす。
つまり、.claude/state/codex-loop/ は「この project の Harness loop state」であり、Codex native state 全体の固定保存先ではない。
状態ファイル
以下はすべて project root 基準。
.claude/state/codex-loop/run.json.claude/state/codex-loop/cycles.jsonl.claude/state/codex-loop/runner.log.claude/state/codex-loop/current-job.json.claude/state/codex-loop/jobs/*.json.claude/state/codex-loop/jobs/*.log.claude/state/codex-loop/jobs/*.out.claude/state/advisor/history.jsonl.claude/state/advisor/last-request.json.claude/state/advisor/last-response.json.claude/state/locks/codex-loop.lock.d
Advisor Consult
Advisor は「代わりに実装する役」ではなく、「次の一手だけ返す相談役」。 loop では次の 3 箇所でだけ呼ぶ。
| タイミング | reason_code | 何をするか |
|---|---|---|
| 高リスク task の初回実行前 | high-risk-preflight | 先に固める観点を聞く |
| 同じ原因の 2 回目失敗後 | retry-threshold | 方針変更か局所修正かを聞く |
| plateau による停止直前 | plateau-pre-escalation | 本当に止めるべきかを聞く |
decision は 3 種だけ。
| decision | loop の扱い |
|---|---|
PLAN | advice を次の executor prompt 先頭に足して再実行 |
CORRECTION | 局所修正の指示として再実行 |
STOP | loop を停止し、理由を state と runner.log に残す |
同じ trigger は trigger_hash = task_id + reason_code + normalized_error_signature で 1 回だけ相談する。
相談回数は task ごとに最大 3 回で、それ以上はユーザー判断に上げる。
注意点
- これは 本当に裏で動く。説明だけ返して終わるスキルではない。
- 同時に 2 本は起動できない。既に走っている場合は
already runningで止まる。 - 既定 executor は Breezing。旧来の 1 task per cycle 挙動が必要な時だけ
--executor taskを使う。 - 失敗したタスクを無理に飛ばして次へ進めるのではなく、基本はその場で止まって理由を残す。
statusとrunner.logを見れば、今どこで止まっているか追いやすい。
具体例
「Phase 41 の残タスクを、今日の間は自動で回したい」なら:
harness codex-loop start 41.1-41.4 --max-cycles 8 --max-workers max --pacing worker途中で様子を見る:
harness codex-loop status夜になって止めたい:
harness codex-loop stopなぜこの形か
Codex では Claude の /loop と同じ wake-up 機構をそのまま使えない。
その代わり、Codex loop runner を土台にして、
Harness 側で状態管理と再入制御を持ち、実作業は Breezing の batch 実行に寄せる。
そうすると、長時間タスクでも「止める」「再開する」「今の状態を見る」が素直になり、
依存関係を満たした作業だけを安全にまとめて進められる。
© Chachamaru127, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
Files
Just SKILL.md in skills-codex/harness-loop of Chachamaru127/claude-code-harness.
Open the folder on GitHubat commit 2b2b748
Compare with similar skills
Harness Loop next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Harness Loop this skillChachamaru127/claude-code-harness | 3.2k | — | ~2.2k | Automated safety check: Notes | MIT | |
| agtx One-Shot Project Runnerfynnfluegge/agtx | 1.7k | — | ~3.8k | Automated safety check: Pass | Apache-2.0 | |
| Cursor Orchestratecursor/plugins | 10k | — | ~1.1k | Automated safety check: Pass | None | |
| GSD Quick Batchopen-gsd/gsd-core | 10k | — | ~1.4k | Automated safety check: Notes | MIT | |
| AutopilotYeachan-Heo/oh-my-claudecode | 40k | 1 repos | ~4.4k | Automated safety check: Pass | MIT | |
| MemPalace Task HandoffMemPalace/mempalace | 59k | — | ~1.9k | Automated safety check: Pass | MIT |
Similar skills
- Official
Cursor Orchestrate
cursor/plugins
Splits a large goal into a tree of parallel Cursor cloud agents, with planners, workers and verifiers coordinated by a script and reporting through structured handoffs.
GSD Quick Batch
open-gsd/gsd-core
Runs several small /gsd-quick tasks as one batch, with a single coordinator that plans them, dispatches workers per item and owns all shared writes and worktree merges.
Autopilot
Yeachan-Heo/oh-my-claudecode
Takes a short product idea through requirements, design, planning, parallel implementation, QA cycles and multi-reviewer validation to produce working code.
MemPalace Task Handoff
MemPalace/mempalace
Creates, hands off, claims, executes and closes agent tasks through the MemPalace logstream, with approval of the exact task before it is recorded.
Zeroshot
the-open-engine/zeroshot
Use Zeroshot to prepare, run, observe, or troubleshoot explicit multi-agent software work locally or on Zeroshot Cloud.
More from Chachamaru127/claude-code-harness
All 25 skills in this repoCI Failure Triage and Repair
Chachamaru127/claude-code-harness
Diagnoses failing CI pipelines and tests, deciding first whether the test or the implementation is at fault, and hands hard cases to a dedicated fixer subagent.
Cursor Composer Task Delegate
Chachamaru127/claude-code-harness
Hands one implementation task to Cursor Composer in an isolated git worktree, then reviews its diff and cherry-picks the result into the main branch.
Acceptance Demo Generator
Chachamaru127/claude-code-harness
Renders a single HTML page showing each acceptance criterion as verified or not, with a ship, wait, or reject recommendation for non-engineers.
Harness Long-Running Task Loop
Chachamaru127/claude-code-harness
Repeats a long task as a series of scheduled wake-ups, each re-entering with fresh context and calling harness-work for one task per cycle.
Harness Plan
Chachamaru127/claude-code-harness
Creates and maintains Plans.md task plans with a spec delta, updates task markers and syncs plan progress with the implementation.
Harness Release
Chachamaru127/claude-code-harness
Runs a release for any project that keeps a Keep a Changelog file on GitHub, from version bump to merge, tag and GitHub Release after a single approval.
Related
Categories
Questions about Harness Loop
What does Harness Loop do?
Starts a background runner that works through ready batches of tasks from Plans.md one cycle at a time via Breezing, with status and stop controls. The skill launches a real background process rather than a simulated loop. Each cycle picks one ready batch, meaning tasks marked cc:TODO or cc:WIP whose dependencies are met, hands it to Breezing, checks the result and moves to the next batch.
When should I use Harness Loop?
Harness Loop fits situations like: running many independent plan tasks unattended over a long session; limiting an autonomous run to a range of Plans.md task IDs; checking the progress of a running background loop; stopping a loop that is partway through its batches.
How do I install Harness Loop in Claude Code?
Run `npx skills add Chachamaru127/claude-code-harness --skill harness-loop -a claude-code`. Or copy the skill folder (skills-codex/harness-loop in Chachamaru127/claude-code-harness) into .claude/skills/harness-loop in your project. Claude Code loads it when a task matches its description.
How do I install Harness Loop in Codex?
Run `npx skills add Chachamaru127/claude-code-harness --skill harness-loop -a codex`. Or copy the skill folder (skills-codex/harness-loop in Chachamaru127/claude-code-harness) into .agents/skills/harness-loop in your project. Codex loads it when a task matches its description.
Can I use Harness Loop in Cursor, Gemini CLI or GitHub Copilot?
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Chachamaru127/claude-code-harness --skill harness-loop -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/harness-loop, .gemini/skills/harness-loop, .github/skills/harness-loop and .opencode/skills/harness-loop in your project.
What does Harness Loop need to run?
Going by SKILL.md and its folder, Harness Loop needs the command-line tools its instructions call (codex and bash). Our summary lists: The harness command-line tool; A Plans.md file listing tasks and their dependencies. Its frontmatter pre-approves these tools: Read, Bash.
Does Harness Loop access the network?
SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.
Is Harness Loop safe to install?
Our automated static check of SKILL.md found notes only (pre-approves every shell command (allowed-tools: bash)), nothing it rates as a warning. It is not a guarantee. Review the folder before installing.
What licence does Harness Loop use?
Harness Loop is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
How many tokens does Harness Loop use?
About 2.2k tokens (SKILL.md is roughly 8.7k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.
What are the alternatives to Harness Loop?
Skills that share tags, products or a category with Harness Loop: agtx One-Shot Project Runner (fynnfluegge/agtx, 1.7k stars), Cursor Orchestrate (cursor/plugins, 10k stars), GSD Quick Batch (open-gsd/gsd-core, 10k stars) and Autopilot (Yeachan-Heo/oh-my-claudecode, 40k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
Who maintains Harness Loop?
Chachamaru127 (a GitHub user) maintains it in Chachamaru127/claude-code-harness, which has 3,154 GitHub stars. The repository holds 25 skills in this directory. The repository was last updated on October 5, 2026.
Source: Chachamaru127/claude-code-harness on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.