Artificial General Research (AGR)

An autonomous iterative optimization framework that uses a 'fresh-context-per-iteration' loop to optimize measurable code metrics.

triggers:autoresearchAGRauto researchoptimization loopauto optimizeartificial general research

GitHub SKILL.md

What it does\nAGR implements an autonomous optimization loop (the Ralph Loop) for any measurable problem. It spawns fresh agent instances to prevent context degradation, implementing a cycle of: experiment → implement → measure metric → verify guard → keep/discard. It is designed to optimize software metrics like speed, accuracy, or bundle size without breaking core functionality.\n\n## When to use it\nUse this skill when you need to perform exhaustive, autonomous research or optimization on a codebase where success can be quantified by a number (metric) and correctness can be verified by a test (guard).\n\n## What's included\n- References: Templates for benchmark scripts, strategy documents, and results logging.\n- Instructions: Detailed decision logic for keeping or discarding experiments based on variance-aware acceptance and artifact detection.\n- Setup Wizard: A guided process to define metrics, guards, and target files.\n\n## Compatible agents\nSpecifically designed for Claude Code (v2.1.72+) using headless mode (`claude -p`), but architecturally agent-agnostic.

Not yet audited

This skill has not been reviewed by our automated audit pipeline yet.

Information

Repository: artificial-general-research
Stars: 29

Related Skills

Cost Tracker

Monitor agent session costs, set budget alerts, and get actionable token-spend optimizations to keep multi-session workflows within budget.

Audit Agents/Skills/Commands

Run a quantitative quality audit on agents, skills, and commands to check production readiness and get prioritized recommendations.

NotebookLM Research Agent

End-to-end research-to-content pipeline: ingest sources into Google NotebookLM, run deep cited research, generate downloadable artifacts, and produce polished c

draw.io Diagram Skill

Create, edit, lint, and export draw.io diagrams; includes export helpers, SVG linting, and repository QA for clean architecture and icon-based diagrams.

Portfolio Optimization

Select and balance project portfolios using constraint-based mathematical optimization to maximize value under budget, resource, and strategic constraints.

API Development

Guidelines and patterns for building robust REST APIs: correct HTTP status codes, RFC 7807 error responses, validation, sanitization, rate limiting, and central

Harness Health Audit

Assess and score the overall health of a Claude Code harness across architecture, skills coverage, hooks, rules, MCP servers, eval pipelines, and team setup.

Multi‑Agent Ideation

Orchestrates multi-agent ideation sessions (Free Thinker, Grounder, Writer, optional Explorer) to explore concepts, capture source materials, and produce struct

Back to Skills

Artificial General Research (AGR)

from artificial-general-research29

An autonomous iterative optimization framework that uses a 'fresh-context-per-iteration' loop to optimize measurable code metrics.

triggers:autoresearchAGRauto researchoptimization loopauto optimizeartificial general research

GitHub SKILL.md

What it does\nAGR implements an autonomous optimization loop (the Ralph Loop) for any measurable problem. It spawns fresh agent instances to prevent context degradation, implementing a cycle of: experiment → implement → measure metric → verify guard → keep/discard. It is designed to optimize software metrics like speed, accuracy, or bundle size without breaking core functionality.\n\n## When to use it\nUse this skill when you need to perform exhaustive, autonomous research or optimization on a codebase where success can be quantified by a number (metric) and correctness can be verified by a test (guard).\n\n## What's included\n- References: Templates for benchmark scripts, strategy documents, and results logging.\n- Instructions: Detailed decision logic for keeping or discarding experiments based on variance-aware acceptance and artifact detection.\n- Setup Wizard: A guided process to define metrics, guards, and target files.\n\n## Compatible agents\nSpecifically designed for Claude Code (v2.1.72+) using headless mode (`claude -p`), but architecturally agent-agnostic.

Not yet audited

This skill has not been reviewed by our automated audit pipeline yet.

Information

Repository: artificial-general-research
Stars: 29

Related Skills

Cost Tracker

Monitor agent session costs, set budget alerts, and get actionable token-spend optimizations to keep multi-session workflows within budget.

Audit Agents/Skills/Commands

Run a quantitative quality audit on agents, skills, and commands to check production readiness and get prioritized recommendations.

NotebookLM Research Agent

End-to-end research-to-content pipeline: ingest sources into Google NotebookLM, run deep cited research, generate downloadable artifacts, and produce polished c

draw.io Diagram Skill

Create, edit, lint, and export draw.io diagrams; includes export helpers, SVG linting, and repository QA for clean architecture and icon-based diagrams.

Portfolio Optimization

Select and balance project portfolios using constraint-based mathematical optimization to maximize value under budget, resource, and strategic constraints.

API Development

Guidelines and patterns for building robust REST APIs: correct HTTP status codes, RFC 7807 error responses, validation, sanitization, rate limiting, and central

Harness Health Audit

Assess and score the overall health of a Claude Code harness across architecture, skills coverage, hooks, rules, MCP servers, eval pipelines, and team setup.

Multi‑Agent Ideation

Orchestrates multi-agent ideation sessions (Free Thinker, Grounder, Writer, optional Explorer) to explore concepts, capture source materials, and produce struct

Artificial General Research (AGR)

Tags

Not yet audited

Information

Related Skills

Artificial General Research (AGR)

Tags

Not yet audited

Information

Related Skills