claude-org-ja: Skill for Claude Code

.claude/skills/org-start/SKILL.md

org-start is a skill for Claude Code from suisya-systems/claude-org-ja. It costs 105 tokens per session (22,169 once invoked), scanned A, original, MIT.

A startup procedure for restoring the organisation's previous state and launching its dispatcher, the session that coordinates work. It expects the coordinator to be running inside a specific Renga operations layout.

In plain words
What is it for?
It helps initialise the coordinator, restore state, launch the dispatcher, and prepare pane communication and role identities.
Why use it?
It brings the organisation back after Claude Code starts without launching duplicate workers or a permanently running curator. The curator is started only when later checks show it is needed.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter. Also seen: reads .claude/ paths; mentions CLAUDE.md; mentions Claude Code.

This is suisya-systems/claude-org-ja's own configuration. It tells Claude Code how to work on claude-org-ja itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything claude-org-ja configures →

Needs its repository: it runs a file that does not travel with it, so clone the repository first. The line is Bash(bash tools/journal_append.sh:*).

Reuse

Borrowing it

Nothing to install: this file belongs to suisya-systems/claude-org-ja. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/suisya-systems/claude-org-ja/main/.claude/skills/org-start/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/suisya-systems/claude-org-ja

Made for: Claude Code.

Wrote 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.

agentmods badge for org-start

README.md
[![agentmods](https://agentmods.dev/badge/skills/suisya-systems/claude-org-ja/org-start/github.svg)](https://agentmods.dev/skills/suisya-systems/claude-org-ja/org-start)
Your own site
<a href="https://agentmods.dev/skills/suisya-systems/claude-org-ja/org-start"><img src="https://agentmods.dev/badge/skills/suisya-systems/claude-org-ja/org-start/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for org-start

Your own site · 80×15
<a href="https://agentmods.dev/skills/suisya-systems/claude-org-ja/org-start"><img src="https://agentmods.dev/badge/skills/suisya-systems/claude-org-ja/org-start.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 105 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 22,169 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector warn 7 Sept 2026
SkillSpector: 4 findings, up to high

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

  • high Rogue Agent · line 177
    Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.
    Fix: Prevent the skill from modifying its own code, SKILL.md, or configuration files. Treat skill files as read-only at runtime.
  • medium Rogue Agent · line 261
    Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
    Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
  • medium Agent Snooping · line 394
    Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
    Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
  • medium Agent Snooping · line 437
    Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
    Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
How audits are shown
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00105 $0.22169
Opus 5 $0.00053 $0.11084
Sonnet 5 $0.00021 $0.04434
Haiku 4.5 $0.00011 $0.02217

Measured 10d ago against content hash d0bd881a3d3e, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

org-start scanned grade A with 1 finding 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 10d ago.

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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

> **sandbox 注記(false negative 回避)**: Claude Code の Bash sandbox はネットワーク / プロセス namespace が隔離されているため、本 Block の常駐起動・疎通確認を sandbox 内で行うと誤判定する。ホスト側で server が port 8099 を正常に listen していても、sandbox 内では `curl` が `000` を返し、`ss` に
.claude/skills/org-start/SKILL.md · 537 lines

How it starts

The opening of the file, as written. The whole thing — 537 lines — stays where its author put it; the contents beside it link to each section on GitHub.

org-start: 組織の起動

ClaudeCode起動後に最初に実行するスキル。前回の状態復元とディスパッチャー起動を行う。

キュレーターは起動しない(オンデマンド化): 常駐キュレーター(spawn + /loop 30m /org-curate)は廃止された。キュレーターはディスパッチャーが worker ペインクローズ時に tools/check_curate_threshold.py の閾値超過を検出したときだけ一時起動される (.dispatcher/references/pane-close.md Step 5)。 org-start では curator_pane_id / curator_peer_idStateWriter.CLEAR で明示クリアする (Block D-5)。curator 不在(null)が正常系である。 worker close が発生しない期間の閾値チェック飢餓ケースへの補助トリガー(org-start バックストップ)は Issue #502 でバックログ化されている。

前提: この Claude は renga --layout ops で起動された窓口ペイン内で動作している。 RENGA_SOCKET / RENGA_PANE_ID 環境変数が継承されているので、mcp__renga-peers__* MCP ツール 14 種(spawn_pane / spawn_claude_pane / close_pane / focus_pane / list_panes / new_tab / send_message / list_peers / set_summary / check_messages / inspect_pane / poll_events / send_keys / set_pane_identity)でペイン操作・ピア通信・画面スクレイプ・lifecycle event 購読・raw キー入力まですべてカバーできる(renga 0.18.0+ で導入された構造。現行 org の renga transport サポート下限は 2.0.0)。 ただしこの 14 種はタブ scope が一様ではないので、主分類 3 つ + 例外で読む(各 capability トークンは独立判定 = 契約 T-§cap の independence rule。分類の正本は docs/contracts/backend-interface-contract.md T-§cap / T-§4.2):

  • Group A(caller_scope)= pane 制御: spawn_pane / spawn_claude_pane / focus_pane / list_panes / inspect_pane / send_keys。この capability を確立できた経路では caller のタブ内 (フォーカス非依存)に解決され、list_panes は caller のタブのみを返す(契約 T-§cap の Group A は spawn_codex_pane を含む 7 種。ここでは org の 14 種に現れる 6 種を挙げている)
  • Group B(caller_scope_close_identity)= 別ゲート: close_pane / set_pane_identitycaller_scope からは導出してはならないclose_pane は不可逆なので特に混同しない)
  • messaging(cross_tab_peers)= タブ横断: send_message / list_peerslist_peers は全タブ列挙、 send_message数値 id 宛ならタブ横断stable name は送信側タブ内でのみ解決する。契約 T-§4.2 は messaging 到達性から pane 制御到達性を推論することを MUST で禁じている
  • 上記 3 分類に入らない 4 種(例外): new_tab は Group A ではなく独立 capability spawn_tab 側(tab 生成 / tab 指定 spawn)。poll_events は pane 制御でも messaging でもなく lifecycle surface で、そのタブ scope は契約 T-§3.1 が別に固定する。set_summary / check_messages は自ペイン / 自キューに 対する操作であって、タブ横断 addressing の話ではない

誤読時の分岐手順(pane_not_found の pane 制御文脈 / messaging 文脈)は .claude/skills/org-delegate/references/renga-error-codes.md を参照。 ここでの 14 種は org が運用で行使する surface であって、プリフライトの必須集合とは別の数である: tools/check_renga_compat.py は capability probe 用の server_info を加えた 15 種を必須(subset 判定、余剰ツールは正常)として検査する。server_info は probe 専用で、 .claude/settings.json の allowlist には意図的に未追加である(理由は docs/design/renga-decoupling.md を参照)。

state DB 前提 (Issue #267 / M4): .state/state.db が唯一の SoT。 read 経路は DB のみ (markdown fallback は M4 で撤去)、構造化セクション (Status / Dispatcher / Curator / Worker Directory Registry / Active Work Items / Resume Instructions) の write 経路は StateWriter.transaction() 経由(post-commit hook が .state/org-state.md を DB から自動再生成、markdown 直接編集禁止 — drift_check が検出する)。free-form ノート(学び / Pending Lead 等) は notes/ 配下に保存する。.state/journal.jsonl は M4 で廃止。 DB が無い場合は python -m tools.state_db.importer --db .state/state.db --rebuild --no-strict で構築する。

輸送層(transport)両系 — 既定 renga / opt-in broker: 本スキルの mcp__renga-peers__* 呼び出しは 既定 rengaORG_TRANSPORT 無設定)で書いてあり、そのまま従えばよい(既定挙動は不変)。ORG_TRANSPORT=broker(opt-in・切戻し可)では MCP サーバー名が org-broker になり、ツールの 完全修飾名が mcp__renga-peers__*mcp__org-broker__* に機械置換される(引数形・セマンティクスは同一なので手順の論理は変わらない)。輸送依存で手順が変わる点だけ broker 併記する:

  • 受信モデル(push 一次 = claude/channel / pull フォールバック): renga は dispatcher / worker のメッセージが <channel source="renga-peers" …> として in-band で push される。broker は push 一次に再設計されている(runtime push-first 0.1.24+、transport-lab docs/design/broker-native-roles.md §9): 各ペイン同居の channel sidecarserver:org-broker-channel)が broker キューを ~1 秒間隔で claim→push し、notifications/claude/channel で本文を idle セッションへ注入する(renga の in-band push と同じ「受けたら即応答」契機)。pull はフォールバック層: sidecar 不在 / unhealthy(heartbeat timeout で delivery_mode=PULL)/ channel 非対応ペイン / claude.ai login 不在時は、各役割が自身の cadence で能動的に check_messages(broker では mcp__org-broker__check_messages)する(§9.6 読み替え表の役割別 cadence。ナッジが出れば契機になりうるが idle を起こさないため能動 poll が受信の正路。既存の「ナッジを見たら check_messages」prose は撤回せずこの fallback cadence として読む。ツール名以外の手順は同型)。契約面は Surface 8 で push 一次が ratified 済み(2026-06-15、S3。broker pull-only を supersede し pull は fallback として retain・renga 不変)。
  • spawn 儀式(folder-trust 承認 + dev-channel sidecar 承認の再導入): renga の spawn_claude_pane--dangerously-load-development-channels server:renga-peers を注入し「Load development channel?」を Enter 承認する(Block D-1)。broker は --mcp-config <broker> を注入し承認プロンプトが Claude Code の folder-trust プロンプトに変わる(send_keys(enter=true) で機械承認)のに加えて、push 一次のため channel sidecar を --dangerously-load-development-channels server:org-broker-channel で load し dev-channel 承認プロンプト(spawn-flow 3-3b)を send_keys(enter=true)再導入機械承認する(ratified §5/§8.5 の folder-trust フローへの加算であり置換ではない。設計は broker-native-roles.md §9.5)。
  • エラー分岐(broker 追加コード): renga コード([split_refused] / [pane_not_found] / [cwd_invalid] 等)に加え broker は [token_invalid] / [session_invalid] / [tool_not_authorized] / [no_backend](= adapter_unavailable)/ [nudge_failed] / [peer_not_found] / [name_taken] を返しうる(未知コードは default-branch で escalate)。一覧は .claude/skills/org-delegate/references/renga-error-codes.md の broker 節を参照。

new_tab / focus_pane は broker surface に無い(意図的除外。本フローは元々使わない)。契約面の正本は docs/contracts/backend-interface-contract.md Surface 8(broker auth & delivery、ratified 2026-06-14。push 一次への additive 改訂 S3 が ratified 済み(2026-06-15、「Ratified amendment」節)・既存 ratified 本文不変更)、設計 SoT は transport-lab docs/design/broker-native-roles.md §9(push 一次再設計)/ docs/design/ja-migration-plan.md §5.2(ii)・§8。broker 実走(dogfood)は Epic #6 Issue G スコープで本スキルの既定経路ではない。

既定の二フレーム注記(Refs #604): ここでの「既定 renga」は運用既定(broker 実走 dogfood が Epic #6 Issue G まで未活性)の意。別にコード既定として tools/transport.py: DEFAULT_TRANSPORT が runtime 0.1.28 (Epic #586) で broker にフリップ済みで、ja 生成器・transport.resolve() はこのコードフレームで render するため生成面は「既定 broker」と表示する — 両フレームは指す対象(運用経路 vs コード定数)が異なり矛盾しない。総説は root CLAUDE.md「輸送層(transport)両系」節。)

Read the full file on GitHub · 537 lines

Changes

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.

  1. 10d ago First seen · 537 lines · 105 tokens per session scan A d0bd881a3d3e

Subscribe to this mod's changes

org-start is a skill published in the GitHub repository suisya-systems/claude-org-ja (5 stars, last pushed 4d ago), licensed MIT. It adds 105 tokens to every session and 22,169 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

orch

AI agent orchestrator — manage teams of AI agents that work on your codebase in parallel. Use when the user wants to: run multiple agents, coordinate AI work, deploy agent teams, manage tasks/goals/agents, check orchestrator status, or mentions 'orch', 'orchestry', 'agents team', 'agent orchestration'.

oxgeneral/ORCH · 71 tokens

board

Read and write the workspace BOARD.md (shared task board for agents + Val). Wraps bin/board, a small Python CLI that preserves the markdown schema and appends to the activity log automatically. Use this whenever an agent proposes, claims, or finishes work that should be visible to the rest of the roster.

OpenCoven/coven · 66 tokens

ai-agent-job-marketplace

The job marketplace where bots hire bots. Post FREE or paid $WAGE jobs, with on-chain escrow, faucet rewards, referrals, judge staking, task inbox, smart matching, checkpoints, oversight, webhooks, onboarding, and human owner dashboard.

rondoflow/rondoflow · 56 tokens

agent-freelance-bounty-marketplace

Earn USDC completing bounties, post jobs, join multi-agent raids, build reputation, rank up. AI agent freelance marketplace with x402 escrow on Solana. Free SOL airdrop on signup. Guilds, ranks, vouching, disputes, Castle Town, leaderboard.

rondoflow/rondoflow · 68 tokens

agent-job-marketplace-with-bidding

Agent-to-agent marketplace MVP. Agents post jobs, bid, contract, submit deliverables, and leave reviews. Payments are off-platform (crypto) in v1.

rondoflow/rondoflow · 41 tokens

dos-dispatch-loop

Run recurring dos-dispatch cycles, switching to dos-replan when the backlog drains and stopping on the kernel's loop verdict. Use for unattended dispatch->replan->dispatch work across disjoint lanes.

anthony-chaudhary/dos-kernel · 49 tokens