Skip to content
TakoAPI
Back to skills

agentdojo

by musashi94

Coding Agents & IDEs
295Downloads reported by clawskills.sh
0Views
0 likesSign in to like this skill
—Not yet rated
Your rating

Brief

Daily low-token, safety-first upskilling loop for OpenClaw multi-agent teams.

Description

What This Skill Does

AgentDojo runs daily micro-drills for AI agent teams, scores output quality, and generates a compact digest. It enforces hard token budgets and safety guardrails across all sessions. Schedule and drill intensity are configured via YAML files.

It combines token-cost enforcement, safety guardrails, and quality scoring into one automated loop instead of requiring separate tools for each concern.

When to Use It

  • Running nightly quality checks on AI agent outputs
  • Tracking skill gaps across agent roles over time
  • Enforcing token budgets during automated training runs
  • Auditing unsafe or policy-override attempts in sourced content
  • Generating daily improvement summaries for multi-agent teams

Example Workflow

INPUT

User asks: Run today's agent drills and give me a summary

AGENT
  1. 1

    Load config/agentdojo.config.yaml and enforce budget, run count, and tool limits

  2. 2

    Select drills from config/drills/*.yaml based on role rotation and recent score gaps

  3. 3

    Execute selected drills in isolated sessions

  4. 4

    Score outputs against docs/scoring-rubric.md and log any audit events

  5. 5

    Save run record JSON and daily markdown summary

OUTPUT

Compact daily digest covering Kurzfazit, new skills learned, a concrete improvement for tomorrow, risks, and next steps

Installation

clawhub install musashi94/agentdojo

Run the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.

Links