Where does Pathfinding save its results?

After reaching Level 5, the skill suggests saving the plan to a standardized location like CLAUDE.md, an ADR (Architectural Decision Record), or specific project directories like .agents/plans/.

When should I trigger the Pathfinding skill?

You should use Pathfinding whenever requirements are unclear, you are exploring a new idea, or you need to plan a complex feature. You can trigger it explicitly using the --pathfind flag or by mentioning 'brainstorming' or 'clarifying'.

What is the confidence scale?

The skill uses a 0-5 bar to track how ready a task is for implementation. Level 0 means you are just gathering context, while Level 5 means the path is fully mapped and you are ready to deliver the final artifact.

How does the skeptic agent help?

The skeptic agent is an automated sanity check that triggers when high complexity or 'gold-plating' is detected. It challenges assumptions to ensure you are following YAGNI (You Ain't Gonna Need It) and DRY principles.

Pathfinding

Name: Pathfinding
Author: outfitter-dev

byoutfitter-dev

Productivity & Workflow

Guides developers through ambiguous requirements using a structured Q&A framework to transform vague ideas into actionable project plans.

About

Pathfinding is a specialized Claude Code skill designed to resolve uncertainty during the early stages of software development. It uses an adaptive, confidence-based Q&A system to scout requirements, explore technical approaches, and clarify unknowns before a single line of code is written. By moving through five distinct levels of clarity, the skill ensures that developers don't waste time on poorly defined tasks. It includes sophisticated features like a 'pushback' protocol for architectural conflicts and an integrated 'skeptic' agent to identify over-engineering and hidden risks, ultimately delivering a clear path forward and persistent documentation.

Key Features

Automated persistence of plans to CLAUDE.md or project documentation folders
Dynamic confidence scoring (0-5) to measure requirement readiness
Built-in 'skeptic' agent to detect and challenge over-engineering
0 GitHub stars
Multi-phase workflow from initial scouting to final artifact delivery
Interactive Q&A using EnterPlanMode for structured decision-making

Use Cases

Mapping out complex refactoring tasks where dependencies are not immediately clear
Brainstorming and evaluating multiple architectural approaches for greenfield projects
Clarifying vague feature requests or ambiguous project requirements

About

Key Features

Automated persistence of plans to CLAUDE.md or project documentation folders
Dynamic confidence scoring (0-5) to measure requirement readiness
Built-in 'skeptic' agent to detect and challenge over-engineering
0 GitHub stars
Multi-phase workflow from initial scouting to final artifact delivery
Interactive Q&A using EnterPlanMode for structured decision-making

Use Cases

Mapping out complex refactoring tasks where dependencies are not immediately clear
Brainstorming and evaluating multiple architectural approaches for greenfield projects
Clarifying vague feature requests or ambiguous project requirements