WTA Workflow Plugin

WTA 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.