# Fadeno: A Repository-Native Playbook Framework for AI Programming Agents

> Fadeno is an open-source YAML Playbook framework designed to address the inconsistency issue of AI programming agents. By defining a workflow once and reusing it in each run, developers can ensure agents execute tasks in the same manner while leaving traceable audit trails.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-06-01T01:46:41.000Z
- 最近活动: 2026-06-01T01:50:15.460Z
- 热度: 161.9
- 关键词: AI编程代理, Playbook, 工作流自动化, 代码审查, Claude Code, Codex, 开发工具, YAML, 开源工具
- 页面链接: https://www.zingnex.cn/en/forum/thread/fadeno-aiplaybook
- Canonical: https://www.zingnex.cn/forum/thread/fadeno-aiplaybook
- Markdown 来源: floors_fallback

---

## Fadeno: A Repository-Native Playbook Framework for AI Programming Agents

Fadeno is an open-source YAML Playbook framework designed to resolve the inconsistency issue of AI programming agents. Its core idea is to define a workflow once, allowing any agent to run in the same way while leaving auditable traces. Original author/maintainer: CrocSwap, Source platform: GitHub, Original link: https://github.com/CrocSwap/fadeno, Release date: June 1, 2026.

## Background: The Consistency Pain Point of AI Programming Agents

With the popularity of AI programming agents like GitHub Copilot, Claude Code, and Codex, developers face the problem of inconsistent behavior: each time they perform a task, they need to repeatedly input similar instructions (e.g., "make a plan, check boundaries, run tests"), and there is no trace of execution records, leading to low efficiency and unreliable results.

## Fadeno's Core Architecture and Design Philosophy

Fadeno (derived from the Esperanto word "thread") adopts a three-layer architecture: 1. Capability Layer: Includes runner/builder skills, role sub-agents, and CLI tools; 2. Definitions Layer: Playbook content is stored in the .fadeno/ directory of the repository; 3. Traces Layer: Execution records are stored in the .fadeno/runs/ directory. The core principle is structured judgment to ensure repeatability and auditability.

## Playbook Structure and Execution Flow

Playbooks are defined in YAML format, containing clear steps and decision gates. Example flow (code-change-review): plan→implement→review→review_gate→test→test_gate→finalize (if failed, revise, maximum 1 loop). Key rules: Gates are judged based on structured outputs (e.g., checking blocking issues in review-report.json), not subjective LLM judgments; loops are bounded and outputs are versioned.

## Cross-Platform Support and Plugin Mechanism

Fadeno is platform-agnostic; the same Playbook can run on Codex and Claude Code. Claude Code users can install via plugins: /plugin marketplace add <owner>/fadeno → /plugin install fadeno@fadeno, then use /fadeno:runner or /fadeno:builder commands. Codex users run npx fadeno init --codex to generate relevant files.

## Layered Enforcement Strategy

Fadeno uses a layered enforcement approach: 1. Advisory Layer: In hosts that only support instructions, request the model to comply with the strategy; 2. Mandatory Layer: Run the fadeno gate command via git/CI/pre-commit hooks, calculate gate conditions based on structured outputs to ensure strict enforcement.

## Use Cases and Value

Fadeno is suitable for: Team collaboration standardization (unified processes), auditable AI development (complete execution traces), cross-platform migration (reducing lock-in risk), and reliable execution of complex tasks (structured steps and gates).

## Quick Start and Future Outlook

Quick Start: Codex users run npx fadeno init --codex; Claude Code users run npx fadeno init --claude. The init command is safe and repeatable; --force overwrites existing files, --with-hooks generates CI/pre-commit guards. Fadeno plans to evolve into an orchestration runtime in the future; the current file-driven design already provides powerful basic functions.