skills.rest
Gamezar
GamezarCOMMUNITY
Explore all skills by this author →
GitHub

sadd-do-and-judge

Orchestrates task execution with AI sub-agents and LLM-as-a-judge verification and automatic retry loops.

Updated Mar 10, 2026
Search Tags:#automation#verification#agent#task execution#llm-as-a-judge#retry loop
One-click install
npx skills add https://github.com/Gamezar/opencode-cek --skill sadd-do-and-judge-gamezar
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: sadd-do-and-judge
Source: https://github.com/Gamezar/opencode-cek/tree/main/plugins/sadd/skills/sadd-do-and-judge
Command: npx skills add https://github.com/Gamezar/opencode-cek --skill sadd-do-and-judge-gamezar

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

This Skill automates the execution of tasks by using specialized AI agents for implementation and verification, ensuring quality through an iterative feedback loop.

Core Features & Use Cases

  • Automated Task Execution: Dispatches sub-agents to perform tasks based on user requirements.
  • LLM-as-a-Judge Verification: Employs an independent AI agent to evaluate the quality and correctness of the implemented task.
  • Iterative Improvement: Automatically retries tasks with feedback from the judge if the initial execution fails, up to a set limit.
  • Use Case: You need to refactor a piece of code. This Skill will assign an agent to do the refactoring, then have another agent review it. If the reviewer finds issues, the refactoring agent will fix them based on the feedback and be reviewed again, repeating until the code meets quality standards or a maximum number of attempts is reached.

Quick Start

Use the sadd-do-and-judge skill to implement a new feature based on the provided requirements.

Frequently Asked Questions about sadd-do-and-judge

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How does LLM-as-a-judge verification work for automated task execution?▼

LLM-as-a-judge verification dispatches an independent AI agent to evaluate the quality and correctness of implemented tasks, parsing its verdict to determine success or trigger feedback-driven retries.

How do I automate code refactoring with AI verification and retries?▼

Automate code refactoring by dispatching implementation sub-agents to perform the task, then using an independent LLM-as-a-judge agent to review the output and automatically retry with feedback until quality standards are met.

What happens when an AI agent fails a task verification?▼

When task verification fails, the system enters an iterative retry loop, providing the judge's feedback to the implementation agent to fix issues, repeating until the task passes or a maximum attempt limit is reached.

Can I use an automated retry loop to handle persistent task execution failures?▼

Yes, the automated retry loop manages persistent failures by feeding verification feedback back to sub-agents, escalating complex issues that remain unresolved after reaching the maximum retry limit.

Does automated task execution support model selection for different sub-agents?▼

Yes, automated task execution handles model selection during task analysis and prompt construction, allowing different models to be dispatched for implementation sub-agents and the independent judge.

What are the limitations of using a retry loop for AI task verification?▼

The retry loop is limited by a set maximum number of attempts; if persistent failures occur due to complex or unresolvable issues, the system will escalate rather than retrying indefinitely.

Related Skills

Explore More Software Engineering Skills→
paralleldriveparalleldrive
377

aidd-parallel

Generate delegation prompts and dispatch sub-agent tasks in dependency order.

Official
Advanced
AGMO-IncAGMO-Inc
4

ralph

Execute persistent task loops with retries, verification, and debugging escalation.

Official
Advanced
tiancaiamaotiancaiamao
5

worker-judge

Coordinate an independent worker-and-judge review loop until task output is approved.

Community
Advanced
dalawwadalawwa

sadd:do-and-judge

Coordinate task execution by dispatching an implementation sub-agent and verifying with an independent judge.

Community
Advanced