What happens once an experiment is finished?

Claude provides options to extract the learnings into formal specifications, create a structured implementation plan, or act on generated follow-up tasks.

What are 'invariants' in Eidos?

Invariants are rules or constraints discovered during the experiment that must remain true in subsequent steps to ensure a successful outcome.

How does this skill track progress?

The skill creates a dedicated Markdown file in a memory directory and uses an intent-try-outcome pattern to document every step taken during the exploration.

What is an Eidos Experiment?

It is a specialized workflow for Claude Code that uses structured logging and phases to explore coding problems where the solution isn't known upfront.

Can I stop and restart an experiment later?

Yes, using the '/eidos:experiment continue' command, Claude can scan your memory files and resume the most recent or specific active experiment.

Eidos Experiment Mode

Name: Eidos Experiment Mode
Author: agenticnotetaking

byagenticnotetaking

•

Productivity & Workflow

Facilitates iterative, log-based exploration for complex coding tasks where the final implementation path is unknown.

Eidos Experiment is a specialized workflow skill for Claude Code designed for high-uncertainty tasks like UI prototyping, deep debugging, and architectural spikes. It implements a rigorous 'intent-try-outcome' loop, maintaining a detailed experiment log that captures emergent invariants and key discoveries in real-time. By treating the log as the primary artifact, it ensures that every trial—successful or not—contributes to a final, mineable specification, effectively bridging the gap between a vague initial intent and a concrete implementation plan.

Key Features

01Emergent invariant tracking to maintain discovered constraints across phases

02Structured intent-try-outcome logging for transparent iterative discovery

03Automatic creation and version-control of timestamped experiment files

0436 GitHub stars

05Automated synthesis of findings into specs, plans, or follow-up items

06Persistent state management with Continue Mode for multi-session tasks

Use Cases

01Exploratory spikes for new technology integrations or architectural refactors

02Systematic root-cause analysis for intermittent bugs or complex system regressions

03Prototyping complex UI/UX interactions where the final 'feel' requires iterative adjustment

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add agenticnotetaking/eidos experiment

For use in Claude.ai and ChatGPT

Download Skill