# wta-workflow-plugin

**Repository Path**: dify_ai/wta-workflow-plugin

## Basic Information

- **Project Name**: wta-workflow-plugin
- **Description**: Claude Code plugin: install and operate the WTA role-gated workflow CLI
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-05-23
- **Last Updated**: 2026-05-26

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# WTA Workflow Plugin

[WTA](https://github.com/langgenius/wta-workflow) is a command-line
tool for teams that share a product repository with both human
engineers and AI coding agents. It records the intent of a change,
breaks it into role-gated tasks, and exports clean commits back to
the public repository, keeping a reviewable `main` branch even
under parallel agent work.

This repository is the Claude Code plugin that teaches AI coding
agents how to install WTA and operate it through the full
Blue-Green role lifecycle. With the plugin loaded, an agent can
walk a user from "I want to use WTA" through project bootstrap,
intent publication, decomposition, implementation, review, and
release.

## Install

In a Claude Code session:

```
/plugin marketplace add langgenius/wta-workflow-plugin
/plugin install wta-workflow@wta-workflow-plugin
```

After install, seven skills become available to the agent.

**Setup and overview:**

- `wta-workflow:wta-install` — clones the WTA source, builds with
  cargo, and places the `wta` binary on `PATH`.
- `wta-workflow:wta-using-wta` — overview of WTA: project setup,
  SSH/VPS bootstrap, identity registration, and a high-level walk
  through the Blue-Green role lifecycle.

**Role-based deep dives** (one per WTA role):

- `wta-workflow:wta-orchestrator` — project setup, intent publish
  and seal, intent close, release push and PR.
- `wta-workflow:wta-green-lead` — decomposition, task creation,
  interface contract signing.
- `wta-workflow:wta-green-impl` — claim, pull worktree, implement,
  run checks, submit.
- `wta-workflow:wta-blue-lead` — acceptance contract signing and
  merge.
- `wta-workflow:wta-blue-harness` — review worktree, run Required
  Checks, record verdict.

The agent triggers each skill automatically when the user's request
matches the skill's description, or you can invoke a skill manually
with `/wta-workflow:<skill-name>`.

Each role-based skill ships with a `reference.md` (schemas and
authority detail) and an `examples/` directory (worked contract
files) so the agent has the contract format and at least one
known-good sample available without cloning the WTA source repo.

## Layout

```
.claude-plugin/
  plugin.json         plugin manifest
  marketplace.json    marketplace listing
skills/
  wta-install/SKILL.md
  wta-using-wta/
    SKILL.md
    reference.md      role x stage x command map
  wta-orchestrator/
    SKILL.md
    reference.md      intent schema, release flags
    examples/intent.md
  wta-green-lead/
    SKILL.md
    reference.md      decomposition, task, interface schemas
    examples/decomposition.md
    examples/task.md
    examples/interface.md
  wta-green-impl/
    SKILL.md
    reference.md      worktree layout, commit identity, submit
    examples/commit-message.md
  wta-blue-lead/
    SKILL.md
    reference.md      acceptance schema, merge gate, verdict types
    examples/acceptance.md
  wta-blue-harness/
    SKILL.md
    reference.md      verdict semantics, review note guidance
```

The structure follows the Claude Code / Agent Skills best-practice
pattern of progressive disclosure: SKILL.md is the high-level
playbook (under 500 lines per the spec), `reference.md` carries the
strict schema details an agent reads on demand, and `examples/`
holds vetted contract files an agent can copy and adapt.

## What the plugin does not do

This plugin does not embed the WTA binary. The `wta-install` skill
walks the agent through `cargo install --path . --locked` against the
public source repository at `https://github.com/langgenius/wta-workflow`.
That repository must be reachable from the user's machine
(authenticated GitHub access while the upstream is private; HTTPS
clone once it is public).

## Compatibility

The `/plugin install` flow is Claude Code-specific. The underlying
SKILL.md files in `skills/` follow the open Agent Skills specification
and can be used by other compatible agents (for example by cloning
this repository and pointing the agent at `skills/<name>/SKILL.md`
directly).

## License

MIT. See [LICENSE](LICENSE).