Skip to content
Get packs

ClawWork Economic Survival Protocol

Lets an agent run a simulated work-or-learn cycle to keep a token-cost balance positive

Source Download skill
Repository
HKUDS/ClawWork
Stars
8,565
Licence
MIT
Files
1 file

What it does

This skill sets up a simulated economy for an AI agent. The agent is told that every model call costs money deducted from its balance, that it can earn money by completing evaluated work tasks, and that it can learn to build knowledge for later tasks. Each day it chooses to work or learn, uses tools such as web search or file creation, submits results, and then stops. It also tracks status tiers such as Thriving, Stable, Struggling and Bankrupt.

Great for

  • Developers running or studying the ClawWork benchmark or economic-agent simulations
  • Researchers evaluating how agents behave under budget and cost pressure
  • Teams building agents that need a structured daily work/learn loop with defined tools

Skip it if

  • General productivity use outside the ClawWork simulation, because the rules and tool names only make sense inside that environment
  • Real-money or production workflows: the payment, balance and earnings described are part of a simulation and the skill itself does not handle real payments, but users might mistake the framing for real financial guidance
  • Environments without the ClawWork tools (decide_activity, submit_work, learn, get_status), where the agent would call tools that do not exist

Try asking

  • “Start today's ClawWork cycle and decide whether to work or learn”
  • “Complete a GDP validation work task and submit the finished file for evaluation”
  • “Create a CSV or Word artifact for a work task and submit it”
  • “Save a research summary to persistent memory with the learn tool”
Step-by-step guide: how to use ClawWork Economic Survival ProtocolInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    skill.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip skill.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip skill.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Use this only inside the ClawWork harness; the tool names must exist or the workflow will fail
  • Review what the agent submits and what it spends on API calls, since the skill instructs it to minimise cost and act quickly
  • The file marks itself always: true, so it is loaded in every session; consider whether you want that for your agent

What's inside

  • SKILL.md

Redistributed unmodified from HKUDS/ClawWork under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.