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/xagi-lab/melra/claude-mdgit clone --depth 1 https://github.com/XAGI-Lab/melraWrote 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/xagi-lab/melra/claude-md)<a href="https://agentmods.dev/instructions/xagi-lab/melra/claude-md"><img src="https://agentmods.dev/badge/instructions/xagi-lab/melra/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 | $0.03392 | $0.03392 |
| Opus 5 | $0.01696 | $0.01696 |
| Sonnet 5 | $0.00678 | $0.00678 |
| Haiku 4.5 | $0.00339 | $0.00339 |
Grade A, and why
melra 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 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.
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 — 214 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
What this is
MELRA is an agent-independent autonomy kernel: the layer an agent asks to
change the world through. Three layers, and the split is the whole design — the
LLM reasons, the harness manages the loop, MELRA owns the effect lifecycle.
MELRA begins where the tool call leaves the model loop. It owns effects, never
reasoning — nothing in packages/ may call a model, and no decision in the
execution path may depend on model output.
The feature test. Before adding anything, ask: would this feature still make sense if the effect request came from ordinary deterministic software rather than an LLM? Authorization, idempotency, credential isolation, recovery, verification, effect history, capabilities — yes, they belong here. Prompt optimization, LLM memory, model selection, a planner, agent personality — no, they belong to the harness above.
For every effect it does exactly nine things and nothing else: type it against a strict schema, classify it, authorise it against policy, gate it on an exact approval phrase, record it durably before anything runs, deduplicate it by idempotency key, run it under a budget and cancel signal, verify it against declared evidence, and receipt it. Work that is not one of those nine jobs — a model router, a planner, a prompt library, semantic memory about the user — belongs to the agent above and does not go in this repo.
MCP over stdio is one of several interfaces onto the same runtime (MCP stdio,
MCP over loopback HTTP, CLI, TypeScript SDK, Python SDK, read-only JSON API);
none of them is a shortcut past a stage of the pipeline. Eleven MCP tools sit in
front of five reference effect adapters (files, terminal, browser, computer, http)
plus
operational memory as a kernel service: six task tools (melra_capabilities,
melra_plan, melra_execute, melra_task_status, melra_task_cancel,
melra_receipt) and five durable-workflow tools (melra_workflow_plan,
melra_workflow_advance, melra_workflow_status, melra_workflow_cancel,
melra_workflow_control). pnpm workspace of TypeScript packages (Node 22+, ESM,
strict tsc), plus two Python projects managed by uv (sdk-py,
benchmarks/browser-agent).
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 · 214 lines · 3,392 tokens per session scan A 2e78e6206a42
melra CLAUDE.md is an instructions file published in the GitHub repository XAGI-Lab/melra (8 stars, last pushed 3d ago), licensed Apache-2.0. It adds 3,392 tokens to every session, about $0.0170 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 2026-08-31.
Other instructions, from other repositories
sandbase-harness AGENTS.md
AGENTS.md instructions for sandbaseai/sandbase-harness, covering sandbase harness agent instructions, project identity, repository map, role a: project maintenance and issue workflow.
Sponsio CLAUDE.md
Claude Code instructions for SponsioLabs/Sponsio, covering agent guide for sponsio, what sponsio is, positioning, what to read first and repository map.
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).
Plonk AGENTS.md
AGENTS.md instructions for ostapondo/Plonk, covering agent rules, layout, adding a module, build & verify and code style.
mcp-dockhand copilot-review.instructions.md
Copilot instructions for strausmann/mcp-dockhand: When reviewing pull requests in this repository.
cyxj-groksearch CLAUDE.md
Claude Code instructions for chenyuxiaojin/cyxj-groksearch, covering claude.md, 常用命令 and 结构与坑.