agents-md-template

This repo hands your AI coding agent a rulebook and your CI the checks that enforce it, catching the shortcuts agents take by default — skipped tests, widened permissions, fail-open model output.

A keel for human + AI project management. Markdown as the database, Git as the audit log, Python as the constitution enforcer — decisions never forgotten, AI never overreaches, drafts expire. | 给「人 + AI 协作的项目管理」装一根龙骨。用 Markdown 当数据库，用 Git 当审计日志，用 Python 脚本当宪法执行者 —— 决策不失忆，AI 不越权，草稿会过期。

The operating system for AI-built software: one rulebook, a skill library, project memory on disk, and gates that hold. Copy it before you start.

Single-file operating standard for Claude Code: verification discipline, honest pushback, calibrated autonomy

Compile ordered Markdown policy modules into one deterministic, verifiable AGENTS.md. The target can be global, repository-local, nested, or another explicit path. The runtime has no dependencies, performs no network access, invokes no shell, and uses no LLM.

"code this idea" — turns a plan or idea into a complete, AI-coding-agent-ready docs set (AGENTS.md + linked architecture/decisions/roadmap/product/design docs). Interviews you with grounded recommendations instead of a fixed template. Works with Claude Code, Codex, Cursor, and other AGENTS.md-compatible agents.

Make Codex smarter with persistent project memory, AGENTS.md templates, smart routing, and session continuity for AI coding agents.

One VS Code sidebar for the memory, rules and skills of every AI coding agent — Claude Code, Cursor, Kiro, Antigravity and AGENTS.md. Shared prompt/skill libraries for your team, one-click skill install, and safe recovery when a project folder is renamed.

If you vibe code with AI — here's a way to turn your project into a living knowledge base that the LLM builds and maintains itself, getting richer every chat instead of starting from scratch every time.

Turn the rules in your CLAUDE.md into deterministic checks, and see whether a real Claude Code run respects them

Wire-format specifications for signing AI instruction files — SKILL.md bundles, agent definitions, CLAUDE.md/AGENTS.md - and the scripts that ship with them.

A reference operating framework for coding agents.

A Codex agent workflow that extracts songs from screenshots and finds verified YouTube matches.

Gated, self-validating framework for AI-assisted development: size-classed workflows (Gates 0–5), decisions as ADRs, learnings as AARs, in-repo issues and wiki — all portable Markdown, enforced by schema-first validation in CI. Well, actually — there's an ADR for that.

AI-assisted production workspace: agent-navigable knowledge base, versioned runbooks, pro git discipline. Claude Code native, AGENTS.md compatible.

A service-file convention for projects built with AI agents: docs for the user, the developer, and the agent.

A practical engineering protocol for AI coding agents: plan complex work, stay in scope, verify before completion, and learn from mistakes.

Template for creating repos that are planned to be used with OpenCode agent

🧠 A second brain your agent builds and maintains - plain-Markdown knowledge graphs with derived backlinks, adversarial verification, and a live report-annotation loop. Every app is a replaceable lens.

Англійська з ШІ-агентом, який не забуває твої помилки між сесіями і не хвалить за текст із трьома помилками. Уроки, словник і розбори — звичайні файли в git.