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/caarlosgg/branchpoint/claude-mdgit clone --depth 1 https://github.com/Caarlosgg/branchpointWrote 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/caarlosgg/branchpoint/claude-md)<a href="https://agentmods.dev/instructions/caarlosgg/branchpoint/claude-md"><img src="https://agentmods.dev/badge/instructions/caarlosgg/branchpoint/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 5.1 | $0.05117 | $0.05117 |
| Opus 5 | $0.02559 | $0.02559 |
| Sonnet 5 | $0.01023 | $0.01023 |
| Haiku 4.5 | $0.00512 | $0.00512 |
Grade A, and why
branchpoint CLAUDE.md scanned grade A with 1 finding 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 4d 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.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
`execSync` de `node:child_process` (sin `simple-git` ni otras How it starts
The opening of the file, as written. The whole thing — 332 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Branchpoint
Concepto
Branchpoint es un servidor MCP (Model Context Protocol) local que audita la
rama Git activa del repositorio del usuario y persiste resúmenes de contexto
por rama en .git/branchpoint/.
El problema que resuelve: cuando un agente IA (Claude Code, Cursor, Cline) trabaja en un repo con varias ramas activas, no tiene memoria de qué se decidió o se hizo en cada rama. Esto provoca dos síntomas:
- Alucinación cruzada de ramas: el agente mezcla contexto de código o decisiones de una rama con el trabajo actual en otra.
- Desperdicio de tokens: el agente tiene que re-explorar y re-explicar el estado del proyecto en cada sesión porque no hay memoria persistente ligada a la rama.
Branchpoint expone herramientas MCP que permiten al agente leer y escribir un resumen de contexto asociado a la rama Git activa en el momento, de forma que cambiar de rama cambia automáticamente el contexto relevante.
Modelo de negocio: open-core. El núcleo vive en GitHub como pieza de portfolio técnico. Funcionalidades avanzadas (equipos, sync remoto, etc.) se venderán como versión comercial en Gumroad. Esta decisión de negocio no debe condicionar decisiones técnicas del núcleo open-source salvo que se discuta explícitamente.
Stack decidido
No cambiar estas decisiones sin discutirlo primero.
- Node 22 LTS — runtime.
- TypeScript en modo strict, ESM puro (
"type": "module",module/moduleResolution:NodeNext). @modelcontextprotocol/sdkv1 (actualmente1.29.0, fijada sin^enpackage.json) — transporte stdio. Explícitamente NO la v2 beta.- Zod v4 — validación de esquemas de las tools MCP.
- tsdown como bundler — NO
tsup, que está sin mantener. - pnpm como gestor de paquetes.
- Biome — lint + format (sustituye a ESLint + Prettier).
- Vitest — testing.
@modelcontextprotocol/inspector— herramienta de testing manual del servidor MCP (se instalará/usará víanpxcuando toque probar, no es dependencia del proyecto).
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.
- 4d ago First seen · 332 lines · 5,117 tokens per session scan A ae0057c1be40
branchpoint CLAUDE.md is an instructions file published in the GitHub repository Caarlosgg/branchpoint (0 stars, last pushed 1mo ago), licensed MIT. It adds 5,117 tokens to every session, about $0.0256 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
openrouter-mcp-multimodal AGENTS.md
AGENTS.md instructions for stabgan/openrouter-mcp-multimodal, covering agent instructions, before you ship, releasing (read this before publishing), short version and version files (must all match package.json).
llm-context.py CLAUDE.md
Instructions for cyberchitta/llm-context.py, covering claude.md, working notes (gitignored) and draining the field notes.
seekstone CLAUDE.md
Claude Code instructions for shaqmughal/seekstone, covering claude.md, what this repo is, commands, the harness itself (run after npm install) and architecture.
Plonk AGENTS.md
AGENTS.md instructions for ostapondo/Plonk, covering agent rules, layout, adding a module, build & verify and code style.
cyxj-groksearch CLAUDE.md
Claude Code instructions for chenyuxiaojin/cyxj-groksearch, covering claude.md, 常用命令 and 结构与坑.
mcp-dockhand copilot-review.instructions.md
Copilot instructions for strausmann/mcp-dockhand: When reviewing pull requests in this repository.