Can I add my own custom rules?

Absolutely. After selecting detected constraints, the skill provides a follow-up prompt to enter custom forbidden or encouraged phrases via comma-separated text.

Can I save my guidance for future sessions?

Yes, the skill writes your preferences to a local configuration file and tracks acknowledged constraints so you don't have to reconfigure the same items repeatedly.

How does it know what to forbid?

The skill parses a local constraint scan file (.jsonl) and prioritizes critical and high-severity issues like hardcoded paths or security vulnerabilities for user review.

What is the Ralph Session Guidance skill?

It is a specialized capability for Claude Code that configures the Ralph loop by identifying code constraints and letting users set forbidden or encouraged behaviors for the session.

When does this skill trigger?

It typically runs as Step 1.6 of the /ralph:start command, but can also be invoked whenever a user mentions reconfiguring session guidance or forbidden items.

Ralph Session Guidance

Name: Ralph Session Guidance
Author: terrylica

byterrylica

•

Productivity & Workflow

Configures session-specific guardrails and priorities for the Ralph loop using automated constraint scans and interactive user flows.

Ralph Session Guidance streamlines the initialization of AI development loops by automatically analyzing project constraints and presenting them as actionable configuration choices. It leverages NDJSON scan results to identify critical risks—such as hardcoded paths or architectural violations—and allows developers to designate them as forbidden actions. By also enabling the selection of 'encouraged' tasks like roadmap priorities or performance optimizations, this skill ensures Claude remains perfectly aligned with project standards and session-specific goals while avoiding repetitive manual configuration.

Key Features

01Seamless integration with the /ralph:start onboarding workflow

02Customizable forbidden and encouraged item lists via manual input

03Interactive AskUserQuestion (AUQ) flows for session configuration

04Dynamic risk-based prioritization of forbidden items from constraint scans

058 GitHub stars

06Learned behavior that filters previously acknowledged constraints

Use Cases

01Standardizing guardrails across different development sessions to maintain architectural consistency.

02Directing AI focus toward specific high-priority project milestones like P0 bug fixes.

03Preventing the AI from introducing security risks or hardcoded values detected in previous scans.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add terrylica/cc-skills session-guidance

For use in Claude.ai and ChatGPT

Download Skill