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/midplaneai/midplane/agents-mdgit clone --depth 1 https://github.com/midplaneai/midplaneWhat 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 | $0.02280 | $0.02280 |
| Opus 5 | $0.01140 | $0.01140 |
| Sonnet 5 | $0.00456 | $0.00456 |
| Haiku 4.5 | $0.00228 | $0.00228 |
Grade A, and why
midplane 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 2d 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 — 158 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent operating notes — midplane-cloud
This is the canonical guidance file for AI coding agents working in this
repository, following the AGENTS.md convention. Tool-specific
files (e.g. CLAUDE.md) point here so there is one source of truth. A nested
AGENTS.md in a subdirectory, if present, takes precedence for files in its
subtree.
Monorepo layout (one codebase, two deployables)
- Control plane (repo root:
apps/web,packages/{db,kms,router},infra/telemetry-proxy) — the web app + hosted MCP proxy. Tests: vitest (./node_modules/.bin/vitest run). MIT exceptapps/web/src/ee/(commercial). - Engine (
engine/, grafted viagit subtree) — the MIT query-path engine, shipped as the compiledmidplane/midplaneDocker image. Tests:bun testviabun run test:engine(NEVER a barebun testfrom root — it would sweep in the control-plane vitest files). Engine governance lives underengine/(LICENSE, THREAT_MODEL.md, TELEMETRY.md, CONTRIBUTING.md, SECURITY.md).
One root bun.lock covers everything (linker = "hoisted"). The engine ships two
ways from one source tree: the image is a self-contained bun build --compile
binary (engine/docker/Dockerfile, repo-root context, no node_modules in the
runtime image), and the midplane npm package is a --target=node bundle
(engine/packages/mcp-server/scripts/build-npm.ts). The engine therefore has to
run on both Bun and Node: bun:sqlite and node:sqlite are reached only
through engine/packages/engine/src/audit/sqlite-driver.ts, whose specifier is
computed so neither bundler sees the other runtime's builtin, and no other Bun
global (Bun.*, import.meta.dir, import.meta.main) may reappear in
packages/*/src. bun test cannot catch a violation — the node-compat CI job
builds the package and drives the real bin under Node against a real Postgres
(scripts/smoke-npm-package.mjs), which is what actually pins this. CI: engine-test.yml /
engine-publish.yml (tags engine-v*) are separate from the control-plane
deploy-fly.yml. The engine still ships as its own image, so the version pin
persists — see "OSS image version pin sites" below.
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.
- 2d ago First seen · 158 lines · 2,280 tokens per session scan A 4d32976ccb45
midplane AGENTS.md is an instructions file published in the GitHub repository midplaneai/midplane (4 stars, last pushed 4d ago), licensed MIT. It adds 2,280 tokens to every session, about $0.0114 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
sysknife CLAUDE.md
Instructions for lacs-project/sysknife, covering sysknife operating notes, current focus — ubuntu; gui paused, pre-commit gate, repository workflow and worktree convention.
accessflow CLAUDE.md
Instructions for bablsoft/accessflow, covering claude.md — accessflow agent rules, project at a glance, repository layout, backend and runtime & framework versions.
niuma-engine CLAUDE.md
Instructions for Destined-at-Dawn/niuma-engine, covering startup sequence (every conversation -- mandatory), project rules, r21 -- three-layer file protection (critical), think before act and no blind overwrite (critical).
belay-mcp AGENTS.md
AGENTS.md instructions for Jairogelpi/belay-mcp: These rules bind any AI agent (Claude, Codex, or otherwise) working on this repository. Adapted from docs/plan.md §9.
pattern8 AGENTS.md
Instructions for Aquifer-sea/pattern8, covering agents.md — p8 global instructions (all agents must follow), pattern 8 (p8) — agent behavior constraints, core constraints, five patterns (cannot be skipped) and 1. pipeline.
vector-plugin GEMINI.md
Gemini CLI instructions for pharosone/vector-plugin, covering vector context for gemini cli, when to use mcp tools vs skills, defaults and auth flow on first mcp call.