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/burakdede/lorepack/agents-mdgit clone --depth 1 https://github.com/burakdede/lorepackWhat 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 | $0.03609 | $0.03609 |
| Opus 5 | $0.01804 | $0.01804 |
| Sonnet 5 | $0.00722 | $0.00722 |
| Haiku 4.5 | $0.00361 | $0.00361 |
Grade A, and why
lorepack AGENTS.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 yesterday.
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 — 220 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Lorepack: agent and contributor rules
Read this before starting any work. It applies to every coding agent and human contributor.
1. What Lorepack is (and is not)
Lorepack is a versioned context build system for AI: "Git and Terraform for the context your AI depends on."
It compiles a directory of documents, spreadsheets and project artifacts into an immutable build that can be inspected, diffed, deployed, activated and rolled back, then read by models through MCP, HTTP or a bounded export.
It is not another local RAG server or MCP document-search tool. Retrieval is a runtime capability, not the product. If a change makes Lorepack a better search engine but a worse build system, it is the wrong change.
Non-negotiable design invariants
- The build is the source of truth. Every runtime (local or remote) is a projection of an immutable build. Source files are never the live serving database; a mutable index is never canonical.
- The lifecycle is the product:
plan → build → validate → activate → diff → rollback. Every feature must strengthen one of: build, inspect, connect, deploy. - Determinism. Identical canonical inputs produce an identical build ID on every OS and path. Timestamps, hostnames, absolute paths and physical SQLite bytes never affect identity.
- Immutability. A failed build or deploy can never corrupt the active version. Activation is a pointer change; rollback never recompiles.
- Provenance is mandatory. Every search result, context item and table row carries a
SourceLocator. A result without one is a bug, not a style issue. - Never invent truth.
authority/status/supersedesare user-declared ranking hints. Lorepack must never claim it detected a conflict or decided which document is correct. - Zero-surprise first run. No Python, Docker, compiler toolchain, native add-on, model download, API key or account. Ever. This constrains dependency choices absolutely.
- Structure before models. Preserve hierarchy, sheets, cell ranges and code boundaries. Flattening a spreadsheet into prose is an invalid implementation.
- Ports and adapters.
coreimports no parser, database, MCP, UI or Cloudflare code. Protocols and backends are adapters behind narrow interfaces. No dynamic plugin loader in v0.1. - Read-only at the AI boundary. No model-facing tool may build, deploy, edit sources or execute shell commands.
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.
- yesterday First seen · 220 lines · 3,609 tokens per session scan A d6341d051107
lorepack AGENTS.md is an instructions file published in the GitHub repository burakdede/lorepack (2 stars, last pushed 12d ago), licensed Apache-2.0. It adds 3,609 tokens to every session, about $0.0180 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
cloudflare-workers-nextjs-saas-template AGENTS.md
Instructions for LubomirGeorgiev/cloudflare-workers-nextjs-saas-template, covering project context, vinext, general coding rules, control flow and comments.
zuke AGENTS.md
Instructions for zuke-build/zuke, covering agents.md, using zuke — the api, without guessing, tech stack, typescript 7 / tsgo and coding guidelines (non-negotiable).
flarestarter CLAUDE.md
Instructions for FlareStarter/flarestarter: This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
baerly-storage CLAUDE.md
Toolchain, verification matrix, module map, conventions, anti-patterns. The main agent entry point.
munkel CLAUDE.md
Instructions for limehq/munkel, covering claude.md, monorepo layout, commands, macos development specifics and deploy.
ai-coding-primer CLAUDE.md
Instructions for ammonhaggerty/ai-coding-primer, covering claude.md, repository structure, guidebook chapters, constraints and writing tone.