Does this skill make changes to my code?

No, this is a discovery and analysis skill. It only reads files and logs its findings to a constraint scan file to guide future actions.

Why are five agents used instead of one?

Using five parallel agents allows for faster, specialized scanning of different domains (Memory, Architecture, Research, Testing, and Freedom) without one task bottlenecking the others.

How does this skill handle documentation links?

The skill uses recursive @ link tracing, meaning if a document references another file using the @ prefix, the agents will follow and analyze those files to an unlimited depth.

What is a 'Constraint Discovery' agent?

It is a specialized sub-agent that scans your project to find rules, limits, and patterns—such as 'Do not modify X' or 'Prefer simplicity'—that Claude should follow while coding.

Which project files does it typically scan?

It scans CLAUDE.md, ADRs (Architecture Decision Records), ROADMAP.md, CI/CD workflows, research logs, and project-specific memory directories like .claude/.

Ralph Constraint Discovery

Name: Ralph Constraint Discovery
Author: terrylica

byterrylica

•

Productivity & Workflow

Analyzes project boundaries and limitations by spawning five parallel agents to map architectural, testing, and philosophical constraints.

Constraint Discovery is a high-performance analysis skill for Claude Code designed to identify the 'degrees of freedom' and hidden boundaries within a codebase. By deploying five specialized Explore agents in parallel, it recursively scans project memory files, architectural patterns, historical research logs, testing gaps, and configuration limits. This skill is essential for ensuring that AI-driven development remains within project-defined parameters, preventing architectural regressions, violations of established philosophy, or the repetition of past experimental failures. It centralizes all findings into a standardized JSONL format, providing a comprehensive safety map for autonomous workflows.

Key Features

01Automated JSONL Aggregation: Merges findings from all agents into a single, machine-readable constraint scan file.

02Recursive @ Link Tracing: Follows document references to unlimited depth across markdown and configuration files.

03Multi-Domain Analysis: Evaluates constraints across architecture, testing, project philosophy, and past research sessions.

048 GitHub stars

05Parallel Agent Execution: Spawns five specialized sub-agents simultaneously for rapid project-wide discovery.

06Risk Mitigation: Explicitly identifies 'hard gates' and 'forbidden patterns' to prevent risky or unauthorized code modifications.

Use Cases

01Automating safety audits during the initialization phase of autonomous coding tasks via /ralph:start.

02Onboarding to a complex codebase to quickly understand established patterns and forbidden practices.

03Analyzing project documentation to identify testing gaps and integration risks before making major architectural changes.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add terrylica/cc-skills constraint-discovery

For use in Claude.ai and ChatGPT

Download Skill