تسجيل الدخول
bladnman

do-work

by @bladnman in Productivity
17
2
# Install this skill:
npx skills add bladnman/do-work

Or install specific skill: npx add-skill https://github.com/bladnman/do-work

# Description

Task queue - add requests or process pending work

# SKILL.md

name: do-work
description: Task queue - add requests or process pending work
argument-hint: run | task to capture

Do-Work Skill

A unified entry point for task capture and processing.

Actions:
- do: Capture new tasks/requests → writes to do-work/ folder
- work: Process pending requests → executes the queue

Routing Decision

Step 1: Parse the Input

Examine what follows "do work":

Pattern Example Route
Empty or bare invocation do work → Ask: "Start the work loop?"
Action verbs only do work run, do work go, do work start → work
Descriptive content do work add dark mode, do work [meeting notes] → do

Step 2: Preserve Payload

Critical rule: Never lose the user's content.

If routing is unclear BUT content was provided:
- Default to do (adding a task)
- Hold onto $ARGUMENTS
- If truly ambiguous, ask: "Add this as a request, or start the work loop?"
- User replies with just "add" or "work" → proceed with original content

Action Verbs (→ Work)

These signal "process the queue":
run, go, start, begin, work, process, execute, build, continue, resume

Content Signals (→ Do)

These signal "add a new task":
- Descriptive text beyond a single verb
- Feature requests, bug reports, ideas
- Screenshots or context
- "add", "create", "I need", "we should"

Examples

Routes to Work

  • do work → "Ready to process the queue?" (confirmation)
  • do work run → Starts work action immediately
  • do work go → Starts work action immediately

Routes to Do

  • do work add dark mode → Creates REQ file
  • do work the button is broken → Creates REQ file
  • do work [400 words] → Creates REQ/CONTEXT files

Payload Preservation Rules

When clarification is needed but content was provided:

  1. Do not lose $ARGUMENTS - keep the full payload in context
  2. Ask a simple question: "Add this as a request, or start the work loop?"
  3. Accept minimal replies: User says just "add" or "work"
  4. Proceed with original content: Apply the chosen action to the stored arguments
  5. Never ask the user to re-paste content

This enables a two-phase commit pattern:
1. Capture intent payload
2. Confirm action

Action References

Follow the detailed instructions in:
- do action - Request capture
- work action - Queue processing

# README.md

do-work

A task queue skill for Claude Code. Capture requests fast, process them later.

Installation

npx add-skill bladnman/do-work

Welcome to your new work loop

This skill gives you a two-phase workflow:

  1. Capture: Throw ideas, bugs, and feature requests at Claude as they come up. Each one becomes a structured request file in do-work/.

  2. Process: When you're ready, tell Claude to work. It picks up pending requests one by one, triages complexity, and builds until the queue is empty.

The idea: separate thinking of things from doing things. Capture is instant. Processing is thorough.

Quick usage

Add a request:

do work add dark mode to the settings page

Creates do-work/REQ-001-dark-mode.md

Add multiple at once:

do work the search is slow, also add an export button, and fix the header alignment

Creates three separate request files.

Process the queue:

do work run

Starts the work loop. Claude triages each request by complexity:
- Simple (config changes, small fixes) → straight to implementation
- Medium (clear goal, unknown location) → explore codebase first
- Complex (new features, architectural) → plan, explore, then build

Each completed request gets archived with its implementation notes and a git commit.

How it works

do-work/
├── REQ-001-pending.md      # Queue (pending requests)
├── REQ-002-pending.md
├── working/                 # Currently being processed
│   └── REQ-003-in-progress.md
└── archive/                 # Completed work
    └── REQ-000-done.md

Requests move through the folders as they're processed. The archive preserves the full history: original request, triage decision, exploration notes, and implementation summary.

Designed for Claude Code

This skill is built for Claude Code, which has:
- Sub-agent spawning (Plan, Explore, general-purpose builders)
- File editing and bash access
- Git integration for per-request commits

It may work with other AI coding tools that support similar agent patterns.

The two actions

Do (capture)

Invoked when you provide descriptive content. Optimized for speed:
- Minimal questions - capture what was said, don't interrogate
- Handles simple one-liners and complex multi-feature specs
- For long inputs, creates a context document preserving the full verbatim text
- Checks for duplicates against existing requests

See actions/do.md for the full capture logic.

Work (process)

Invoked when you say "run", "go", "start", or just confirm the prompt. Runs the build loop:
- Triages each request to determine the right amount of planning
- Spawns specialized agents only when needed
- Archives completed work with implementation notes
- Creates atomic git commits per request

See actions/work.md for the full processing logic.

License

MIT

# Related Skills

moltbot
obsidian @moltbot

Work with Obsidian vaults (plain Markdown notes) and automate via obsidian-cli.

★ 65,001

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

⚡ Amp 🚀 Antigravity 🤖 Claude Code 🦀 Clawdbot 📝 Codex ▶️ Cursor 🤖 Droid 💎 Gemini CLI 🐙 GitHub Copilot 🪿 Goose 📊 Kilo Code 🔧 Kiro CLI 💻 OpenCode 🦘 Roo Code 🌲 Trae 🏄 Windsurf

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.