Skip to content
Get packs

Agent Harness Construction

Design an AI agent's tools, outputs and error recovery so it finishes tasks more reliably.

Source Download skill
Repository
affaan-m/ECC
Stars
295,127
Licence
MIT
Files
1 file

What it does

This skill gives an AI agent guidance on how to structure the tools it can call, the format of their results, and how to recover from errors. When you are building or tuning an agent system, the agent will suggest clear tool names, consistent status and summary fields, explicit stop conditions, and a hybrid planning approach. The aim is to help agents complete tasks more often and with fewer retries. It is a set of design principles written in Japanese, not a program that runs anything.

Great for

  • Developers designing tool sets and response formats for their own AI agents
  • Teams diagnosing agents that loop, retry too much, or fail to finish tasks
  • People who want a checklist of error-handling and context-budget practices

Skip it if

  • End users looking for a ready-made automation; it does not perform tasks itself, it only gives design advice
  • Projects needing measured proof: the file lists metrics to track but provides no benchmarks or evidence that the advice works

Try asking

  • “Review my agent's tool list and tell me where it has overlapping or catch-all tools”
  • “Design a consistent response format with status, summary, next_actions and artifacts for my file tools”
  • “Add error recovery rules with root-cause hints and stop conditions to my agent's failure paths”
  • “Suggest whether ReAct or function calling fits my multi-step research agent better”
Step-by-step guide: how to use Agent Harness ConstructionInstall, 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.

    agent-harness-construction.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip agent-harness-construction.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip agent-harness-construction.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

  • Give the agent your actual tool definitions or code so it can apply the rules to concrete names and schemas rather than generic ones.
  • Ask for the metrics listed (completion rate, retries per task, pass@1 and pass@3) to be built into your own evaluation before relying on the design changes.
  • The skill is written in Japanese; ask in English or Japanese, but expect the advice to be framed as general principles that still need adapting to your stack.

What's inside

  • SKILL.md

Redistributed unmodified from affaan-m/ECC 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.