Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/rubenmarcus/ralph-starter/claude-mdgit clone --depth 1 https://github.com/rubenmarcus/ralph-starterWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/rubenmarcus/ralph-starter/claude-md)<a href="https://agentmods.dev/instructions/rubenmarcus/ralph-starter/claude-md"><img src="https://agentmods.dev/badge/instructions/rubenmarcus/ralph-starter/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01239 | $0.01239 |
| Opus 5 | $0.00620 | $0.00620 |
| Sonnet 5 | $0.00248 | $0.00248 |
| Haiku 4.5 | $0.00124 | $0.00124 |
Grade A, and why
ralph-starter CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 119 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
Context for Claude Code when working on ralph-starter.
Package Manager
This repo uses pnpm for development. The packageManager field and a preinstall script enforce this.
Note: ralph-starter itself supports any package manager in user projects via detectPackageManager() in src/utils/package-manager.ts. Do not hardcode pnpm in agent prompts or user-facing code.
What is ralph-starter?
A CLI tool that runs autonomous AI coding loops. It fetches specs from GitHub, Linear, Notion, Figma, and OpenSpec, then orchestrates coding agents to build software automatically. Each loop iteration spawns a fresh agent process with a progressively trimmed context — continuity is maintained through the filesystem (IMPLEMENTATION_PLAN.md, .ralph/iteration-log.md), not conversation history.
Quick Commands
pnpm build # Build TypeScript
pnpm lint # Lint code (Biome)
pnpm lint:fix # Lint and auto-fix
pnpm test # Run tests (Vitest, watch mode)
pnpm test:run # Run tests once
pnpm dev # Development mode (tsc --watch)
pnpm format # Format code (Biome)
pnpm check # Check code (Biome lint + format)
pnpm typecheck # Type-check without emitting
Repository Structure
src/- TypeScript source code (see directory overview below)docs/- Docusaurus documentation sitedist/- Compiled outputspecs/- Spec files for development
Key Concepts
Integrations - Fetch specs from external sources:
- GitHub: Issues, PRs, files
- Linear: Tickets by team/project
- Notion: Pages and databases
- Figma: Design specs and tokens
Agents - AI coding assistants ralph-starter orchestrates:
- Claude Code, Cursor, OpenCode, OpenAI Codex, GitHub Copilot, Gemini CLI, Amp, Openclaw
Loop Executor - Runs agents in autonomous loops until task completion. Each iteration: build context → spawn agent → collect output → run validation (lint/build) → commit → repeat or exit.
LLM Providers - For internal features (Anthropic, OpenAI, OpenRouter)
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 5d ago First seen · 119 lines · 1,239 tokens per session scan A ddbf48bbc091
ralph-starter CLAUDE.md is an instructions file published in the GitHub repository rubenmarcus/ralph-starter (106 stars, last pushed 1mo ago), licensed MIT. It adds 1,239 tokens to every session, about $0.0062 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
agent-orchestrator AGENTS.md
Instructions for geserdugarov/agent-orchestrator, covering repository guide for ai agents, what this project is, where the details live, repository layout and development.
gh-stars-classifier-skill AGENTS.md
Instructions for catuscio/gh-stars-classifier-skill, covering github stars classifier — agent workflow, when to invoke, prerequisites (check first, fail fast), root resolver (run once, before any phase) and workflow.
gh-stars-classifier-skill GEMINI.md
Instructions for catuscio/gh-stars-classifier-skill, covering github stars classifier — gemini cli entry and gemini-specific setup.
githueber AGENTS.md
Instructions for lkoelman/githueber, covering rules, test-driven development (tdd) workflow, tdd steps, editing and change discipline and hard rules.
agent-orchestrator CLAUDE.md
Instructions for geserdugarov/agent-orchestrator, a project described as: Orchestrator for automated GitHub issue monitoring and agentic development across multiple repositories.
apm python.instructions.md
Python development guidelines.