Do And Judge

A solid pattern for when you need quality gates on implementation work. This spins up three agents: a meta-judge to write evaluation criteria, an implementation agent to do the actual work, and a judge to verify against those criteria. The orchestrator (you) just coordinates and never touches code directly, which keeps contexts clean. It runs meta-judge and implementation in parallel, then loops up to twice if the judge fails the work. The rubric generation before judging is smart since generic code review often misses task-specific requirements. Main tradeoff is token cost from running multiple agents, but you get consistent verification and feedback loops without polluting your main context with implementation details.

skills.sh installs: 1.3K
GitHub stars: 1.8K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add neolabhq/context-engineering-kit --skill do-and-judge --agent claude-code

Installs into .claude/skills of the current project.

Files

do-and-judge

Task

Execute a single task by dispatching an implementation sub-agent, verifying with an independent judge, and iterating with feedback until passing or max retries exceeded.

Arguments

Argument	Format	Default	Description
`task`	Free-form text	Required	Task description to execute
`--model`	`haiku\|sonnet\|opus`	auto-selected	Explicit user override for all sub-agents: implementation, meta-judge, and judge. When omitted, you MUST select the model per the Model Selection Policy — there is no fixed fallback tier. When provided, the user's choice wins over the policy for every sub-agent — see the Escalation Rule for how escalation interacts with an explicit override.
`--strict`	`--strict`	`false`	Disable the Iteration Discretion Rule - the task passes ONLY when `score >= 4.0`, otherwise retry until max retries is reached.