Borrowing it
Nothing to install: this file belongs to satoh-y-0323/claude-code-conductor. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/satoh-y-0323/claude-code-conductor/main/.claude/skills/parallel-agents/SKILL.mdgit clone --depth 1 https://github.com/satoh-y-0323/claude-code-conductorWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/satoh-y-0323/claude-code-conductor/parallel-agents)<a href="https://agentmods.dev/skills/satoh-y-0323/claude-code-conductor/parallel-agents"><img src="https://agentmods.dev/badge/skills/satoh-y-0323/claude-code-conductor/parallel-agents.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00052 | $0.12044 |
| Opus 5 | $0.00026 | $0.06022 |
| Sonnet 5 | $0.00010 | $0.02409 |
| Haiku 4.5 | $0.00005 | $0.01204 |
Grade A, and why
parallel-agents scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 421 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Parallel Agents（PO 後継、v1.12.0+）
/develop のフェーズ D で plan-report に YAML フロントマター内の po_plan_version 値が "0.1" のときに Read される手順（"sequential" または フロントマター無しのプランは dev-workflow の legacy TDD 経路へ振り分けられる）。
C3 の親 Claude が plan-report の DAG を wave 単位 で歩く:
- 各 wave 内のタスクを 1 ターン内で複数 Agent ツール並列起動
- 各 Agent は
isolation: "worktree"付きで独立した git worktree 内で完結 - 親 Claude が wave 完了後に各 worktree の成果物を取り込み、一括コミット
permission race の構造的修正（2026-05-11 PoC で 15 並列・101 tool 呼び出し 0 失敗確認）と公式 isolation: "worktree" フロントマターにより、旧 PO（Parallel Orchestra）の並列実行レイヤを Claude Code 公式機能で代替する。
depth 1 制限について
Claude Code のサブエージェントは 更にサブエージェントを spawn できない（公式仕様 depth 1 制限）。v2.2.0 時点で配布されている全 agent は内部で Agent ツールを使わない設計のため、すべて並列起動可能。
将来的に「内部で Agent ツールを使う agent」を追加する場合は、その agent を含む wave のタスク数を 1 に絞る運用ガードが必要になる（v2.0.0 まで存在した tdd-develop agent はこのパターンだった）。
subagent_type 明示指定と wt_* 名前空間（v2.2.0+）
subagent_type パラメータには カスタム agent (.claude/agents/*.md) も指定可能（公式仕様）。これにより frontmatter の permissionMode / tools / model / memory が subagent 起動レイヤーで自動適用される。
並列実行で permission プロンプトに詰まらないよう、v2.2.0 から worktree 専用の wt_* プレフィックス agent を導入した:
wt_tester/wt_developer/wt_systematic-debugger: frontmatter にpermissionMode: bypassPermissionsを持つ並列専用バリアント- 本体ロジックはオリジナルの
tester/developer/systematic-debuggerと同等。差分はレポート出力のファイル名規約のみで、並列専用 agent はtest-report-{task_id}.md/debug-needed-{task_id}.md/debug-analysis-{task_id}.mdを主経路とする（タイムスタンプ形式は task_id 不在時の保険） - worktree 内のみで動作するため、
.claude/hooks/worktree_guard.py(PreToolUse,PO_WORKTREE_GUARD=1) が worktree 外への書き込みをブロックする保護下にある
直接起動経路（worktree なし）では元の agent を使い、main リポジトリでの bypass を防ぐ。
前提条件
- plan-report が
.claude/reports/plan-report-*.mdの形式で配置され、YAML フロントマターを持つこと（po_plan_version: "0.1"） po_plan_version: "sequential"またはフロントマター無しの plan-report は D-0 が legacy 逐次 TDD（D-1〜D-5）へ振り分ける（本 skill には入らない）- Claude Code の Agent ツールが
isolation: "worktree"パラメータをサポートしていること（v2.1.x 以降）。v2.1.150 未満では Agent 完了時の worktree auto-cleanup が動作しないため、2-F-3 のフォールバック手順が毎回必要になる挙動差がある（利用者向け推奨設定は.claude/docs/parallel-agents-setup.mdを参照）
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 421 lines · 52 tokens per session scan A 4dd37d8cd42d
parallel-agents is a skill published in the GitHub repository satoh-y-0323/claude-code-conductor (1 stars, last pushed 2d ago), licensed MIT. It adds 52 tokens to every session and 12,044 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-04.
Other skills, from other repositories
git-workflow
Essential Git patterns for effective version control, eliminating redundant Git guidance per agent.
multi-agent-worktree-discipline
Operate safely in repositories where multiple coding agents (Claude, Codex, Cursor, humans) work in parallel git worktrees. Covers session bootstrap in a fresh worktree, toolchain/version pinning, lockfile-safe dependency installs, branch topology and reconciliation, recovering "lost" work without redoing it, commit…
pr-to-video
Turn a GitHub pull request (a PR URL, owner/repo#N, or 'this PR' in a checked-out repo) into a code-change explainer video — changelog, feature reveal, fix, or refactor walkthrough built from the diff, commits, and files: the input is a code change, not a website. Not a product promo (/product-launch-video) or a no-PR…
agent-framework-py-release
Use when cutting a Python release for the microsoft/agent-framework monorepo. Triggers on "bump py versions", "cut a python release", "prepare release PR for python", "release py packages", "bump python to X.Y.Z", or similar requests to bump Python package versions and prepare a release PR. Handles all four lifecycle…
flow-next-land
Autonomous PR babysitter tick. Fixes CI, resolves feedback, merges when converged, closes the spec, releases. Emits LANDVERDICT. Use when asked to land PRs.
flow-next-worktree-kit
Manage git worktrees (create/list/switch/cleanup) and copy .env files. Use for parallel feature work, isolated review, clean workspace, or when user mentions worktrees.