workflow-toolbox

Description

Toolbox for Claude Code's Workflow tool — author deterministic multi-agent workflow scripts (workflow-composer, with the @workflow-toolbox compile-time pattern toolkit as its standard library), scaffold a build-clean workflow skeleton from a spec (toolkit-scaffold), diagnose a failed run from its journal and decide whether resumeFromRunId helps (workflow-debugger), re-verify the runtime surface after a Claude Code upgrade (upgrade-canary), run a bias-free multi-lens analysis of any subject when relevant (independent-analysis), and run a deep-grounding pass that collects and verifies evidence against the real sources before asserting or analyzing (deep-grounding), and install, check, and migrate editable, versioned copies of the plugin's own rules and agent definitions into a project's or user's config directory (adopt). Ships the independent-analysis workflow that skill triggers (domain-agnostic adversarial multi-lens analysis), a Stop hook that auto-surfaces a finished Workflow run's cost + traceability audit report (always a notice; grabs the session when the run looks like trouble, was degraded by a silently-denied tool call, or requested external delegation whose agents show no real external-CLI invocation), and a tracker-agnostic actionability Stop gate that blocks only while startable work remains with nothing running, paired with a Planka producer that writes the snapshot it reads automatically from board-reading tool calls (never a guess: it writes nothing from a partial read or a read without card descriptions, uses the project's own dependency parser when one exists and the shipped one otherwise, and never counts a card without a Depends-on line as startable). Also ships the pilot agent suite (pilot / pilot-orchestrator / pilot-watchdog) and the pilot-wave composer skill for driving tracked task-tracker cards by delegation, plus a conditional SessionStart hook that injects a generic delegation ladder where a project does tracked/delegated work and points at the delegation-chain skill (a non-user-invocable reference on spawn shapes, the upward report chain, wake signals, and what each shipped watcher/guard does and does not cover) before a delegation decision is made. Ships a Planka kanban skill for any project that wants board-based task tracking: planka-tracking (onboard a project — create the board, migrate an existing progress.md into cards, write the pointer), requiring only a reachable planka MCP server.

What it can do

Reach level 2 — writes or runs What does this mean? →

Raw API calls (9) +

• $.clock.every
• $.command.register
• $.env.get
• $.process.run
• $.ui.close
• $.ui.invalidate
• $.ui.log
• $.ui.open
• $.ui.resolve

Static check passed on Claude Code 2.1.295.

Install

Check the repo for the exact name.

Details

Repository

home-dev-lab/workflow-toolbox

Path in repo

plugin

License

Unknown

Last updated

Oct 9, 2026

First seen

Oct 10, 2026

Last scanned

Oct 10, 2026

Related mods — reach level 2

• ruview-live ★ 98,587
• ruflo-agntcy ★ 76,337
• ruflo-adr ★ 76,337
• ruflo-agent ★ 76,337