talkingclaw AGENTS.md

talkingclaw AGENTS.md is an instructions file for Codex, OpenCode from shibu003/talkingclaw. It costs 943 tokens per session, scanned A, original, Apache-2.0.

A set of repository instructions for TalkingClaw, a system that controls rooms, terminals, files, audio transcription, automation, and documentation.

In plain words
What is it for?
Use it before modifying TalkingClaw to follow its change gates, avoid disrupting running services, inspect shared work, and report completion correctly.
Why use it?
It defines safety rules for making changes alongside another coding agent, including protected processes, required preparation, worktree use, and handoff practices.

Instructions file for CodexOpenCode

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/shibu003/talkingclaw/agents-md
Clone the repo
git clone --depth 1 https://github.com/shibu003/talkingclaw

Made for: Codex, OpenCode.

Per session 943 This file is loaded in full into every session.
When invoked 943 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
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 $0.00943 $0.00943
Opus 5 $0.00472 $0.00472
Sonnet 5 $0.00189 $0.00189
Haiku 4.5 $0.00094 $0.00094

Measured 2d ago against content hash 3c528d7a3281, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

talkingclaw AGENTS.md 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 2d 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.

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.

AGENTS.md · 39 lines

What it actually says

talkingclaw — Codex 向け作業規約

まず CLAUDE.md を読み、その 5 ルールすべてに従うこと(完成の定義・会話最優先・正直・自傷しない・変更の反映)。特に:

  • port 3300(部屋)と port 10101(音声エンジン)のプロセスを止めない
  • test/accept-*.sh を勝手に実行しない(部屋が落ちる)
  • src/room.ts 等サーバ側の変更は部屋の再起動が必要 — 勝手に再起動せず、ユーザーに伝える

並走ルール(Claude Code と同じリポジトリで作業するため)

  1. 作業は別 worktree で: git worktree add ../talkingclaw-codex -b codex/<topic>。main のツリーを直接編集しない
  2. main への取り込みは直列(片方ずつ)。git add -A 禁止、パスを名指しして add する
  3. 着手前に .kaiwa-loop/handoffs/ の最新を読む。着手と完了は handoff への追記か commit で宣言する(沈黙・mtime の停止を完了扱いしない)
  4. ログ・中間ファイルは codex- を冠する(run.log のような一般名は禁止 — 相手の出力を自分の結果として読む事故になる)
  5. 同じ機能を二重実装しない — 書く前に相手の直近成果物を grep する

実装前ゲート(newway)— Codex には hooks が効かないのでこの節が唯一の強制

実装・作成の依頼を受けたら、正本 ~/OhJOJO/newway.md に従うこと。

  • 対象 PBI の G1(受入基準 SBE 表・テスト設計・スコープ外・不確実性表)が揃っていなければ実装を始めない。無ければ §11 テンプレで PBI を起こし §13 P1→P2→P3 を通す
  • backlog は repo 直下の backlog/ に置く
  • done 報告は §12 G2 の表(AC pass/fail + ドリフト判定)を添えてのみ行う
  • 「急ぎ」でも最小形(AC 1 行 + 例外 1 行 + スコープ外 1 行 + 検証 1 つ)未満で実装を開始しない

共有 skill

~/.codex/skills/ に Claude Code の自作 skill 一式を共有済み(正本は ~/.claude/skills/、symlink)。 コマンド(clear-prep / gowave / plantoroad)は ~/.codex/prompts/ にある。この repo で特に重要なもの:

  • safe-commit — 並列作業中の commit 混入防止。commit 前に必ず従う
  • site-api — web から同じ形のものを 3 件以上取る時の唯一の正本
  • data-maestro — 取得したデータを数える・比べる・解釈する前に読む
  • gowave — Wave(0.5〜3h の commit 可能単位)で回す時の不変ルール

注意: commanderclaude -p を、figma-kit は Claude 側の Figma MCP ツールを前提に書かれている。 手順内のツール名が手元に無い場合は、そこだけ読み替えるか着手前にユーザーに確認する。

過去のミス台帳は ~/.claude/LESSONS.md(索引)。行動前に該当する 1 ファイルだけ読む。

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. 2d ago First seen · 39 lines · 943 tokens per session scan A 3c528d7a3281

Subscribe to this mod's changes

talkingclaw AGENTS.md is an instructions file published in the GitHub repository shibu003/talkingclaw (1 stars, last pushed 16d ago), licensed Apache-2.0. It adds 943 tokens to every session, about $0.0047 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 2026-08-31.