charlie947/agent-harness-starter

Scaffold for setting up an agent harness in Claude Code or Cowork. Companion to the MarTech AI newsletter.

22Stars on the repository
6Mods indexed here, across every type
3mo agoLast push, which is what freshness is scored on
MITLicence, which decides whether bodies are shown

builder

01

charlie947/agent-harness-starter

Agent

Executes the approach handed over by the strategist. Ships the actual artifact (file, code, post, design, doc). Does not re-decide what to build. Hands off to qa-scorer.

not rated 22 3mo ago A 44 tokens original MIT

qa-scorer

02

charlie947/agent-harness-starter

Agent

Hard quality gate. Scores the builder's output against a defined rubric. Returns PASS only if the score clears the threshold. Returns a specific list of lost points if below threshold so the builder can fix. Never lets sub-threshold work through to the user.

not rated 22 3mo ago A 56 tokens original MIT

strategist

03

charlie947/agent-harness-starter

Agent

Decides WHAT to build and WHY. The judgment agent at the front of a pipeline. Use when the user has a goal but hasn't decided on the specific approach yet. Reads context, considers options, recommends one with the tradeoff named. Hands off to builder.

not rated 22 3mo ago A 56 tokens original MIT