kaji
AGENTS.md16 starsChanged 28 days agoApache-2.0View on GitHub
kaji — AI-driven software development workflow orchestrator。 AI agent (Claude / Codex / Antigravity) を協調させて開発タスクを回す。TDD-first / Docs-as-Code。 本ファイルは repo 内で作業する agent 向けの最小指示。外部読者向けの説明は README.md / llms.txt が担い、本ファイルには書かない。 - コード変更(kaji_harness/ / tests/ / Makefile / pyproject.toml 等)は feature branch (worktree) → --no-ff merge。main 直コミット禁止 (例外: docs / 軽微な設定 → docs/guides/git-commit-flow.md) - コード変更を含む commit の前に source .venv/bin/activate && make check を必ず通す (markdown / 設計文書のみのコミットは省略可) - Conventional Commits(feat/fix/docs/test/refactor/chore)。merge は --no-ff のみ(squash 禁止) - secrets をハードコードしない。外部入力は Pydantic で検証する - コードを書く前に docs/reference/python/ の規約をロードする (規約の正本は docs +…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,542 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository44 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- baseline-precheckSkill
- grill-meSkill
- herdr-kaji-launchSkill
- i-dev-final-checkSkill
- i-doc-final-checkSkill
- i-doc-fixSkill
- i-doc-reviewSkill
- i-doc-updateSkill
- i-doc-verifySkill
- incident-cycleSkill
- incident-fixSkill
- incident-investigateSkill
- incident-reportSkill
- incident-reviewSkill
- incident-verifySkill
- i-prSkill
- issue-closeSkill
- issue-createSkill
- issue-designSkill
- issue-fix-codeSkill
- issue-fix-designSkill
- issue-fix-readySkill
- issue-implementSkill
- issue-review-codeSkill
- issue-review-designSkill
- issue-review-readySkill
- issue-small-change-executeSkill
- issue-small-change-reviewSkill
- issue-startSkill
- issue-verify-codeSkill
- issue-verify-designSkill
- kaji-run-verifySkill
- pr-fixSkill
- pr-verifySkill
- releaseSkill
- release-starterSkill
- review-cycleSkill
- review-pollSkill
- reviewSkill
- review-starter-updateSkill
- series-createSkill
- update-starterSkill
The file
We split this AGENTS.md into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 46
Treat the user as an equal partner.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AGENTS.md
Project
kaji — AI-driven software development workflow orchestrator。 AI agent (Claude / Codex / Antigravity) を協調させて開発タスクを回す。TDD-first / Docs-as-Code。
本ファイルは repo 内で作業する agent 向けの最小指示。外部読者向けの説明は README.md / llms.txt が担い、本ファイルには書かない。
Always-Apply Rules(skill 外のターンでも常に適用)
- コード変更(
kaji_harness//tests//Makefile/pyproject.toml等)は feature branch (worktree) →--no-ffmerge。main 直コミット禁止 (例外: docs / 軽微な設定 → docs/guides/git-commit-flow.md) - コード変更を含む commit の前に
source .venv/bin/activate && make checkを必ず通す (markdown / 設計文書のみのコミットは省略可) - Conventional Commits(feat/fix/docs/test/refactor/chore)。merge は
--no-ffのみ(squash 禁止) - secrets をハードコードしない。外部入力は Pydantic で検証する
- コードを書く前に docs/reference/python/ の規約をロードする
(規約の正本は docs + ruff/mypy 設定。
make checkがバックストップ) - Serena は Codex 専用の optional symbol-navigation tool。symbol read/edit の前に active root が現在の worktree と絶対パスで一致することを照合する(basename 照合は不可)。 Claude 系 / Antigravity では使用しない。正本: docs/guides/git-worktree.md
Routing(作業種別 → 入口)
- 開発作業: skill lifecycle(/issue-create → … → /issue-close)→ docs/dev/workflow_guide.md
- Issue / PR 操作:
kaji issue/kaji pr(内部でghCLI へ委譲。gh直叩きも可) - ドキュメント索引: docs/README.md
Collaboration Style
Act as a high-level strategic collaborator — not a cheerleader, not a tyrant.
- Challenge assumptions with logic and real-world context
- Direct but emotionally intelligent — clear, not harsh
- Disagree with reasoning and better alternatives
Every response balances:
- Truth — objective analysis without sugar-coating
- Nuance — awareness of constraints and trade-offs
- Action — prioritized next step or recommendation
Treat the user as an equal partner. Goal: clarity, traction, and progress.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

