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/lodyai/lody/agents-mdgit clone --depth 1 https://github.com/LodyAI/LodyWrote 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/lodyai/lody/agents-md)<a href="https://agentmods.dev/instructions/lodyai/lody/agents-md"><img src="https://agentmods.dev/badge/instructions/lodyai/lody/agents-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 | $0.01789 | $0.01789 |
| Opus 5 | $0.00894 | $0.00894 |
| Sonnet 5 | $0.00358 | $0.00358 |
| Haiku 4.5 | $0.00179 | $0.00179 |
Grade A, and why
Lody 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 today.
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 — 131 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Repository guidelines
CLAUDE.md is a symlink to this file. Edit AGENTS.md only.
Context maintenance
Read every AGENTS.md from the repository root to the file being changed.
Before PR or Issue work, read .github/AGENTS.md. Record public contributor
invariants in the narrowest relevant AGENTS.md (under 8 KiB; add a CLAUDE.md
symlink for new scoped files). Internal context, plans, specifications, and
task records stay in the private repository.
Repository boundary
Standalone public source tree: apps/{cli,electron} and the packages they
consume. Excludes hosted backends, operator/billing config, private secrets,
and Web/mobile app sources.
- Never add a dependency on
@lody/convex, a private workspace package, or a generated backend API declaration. - Public optional-cloud protocol names/DTOs live in
packages/cloud-api. - Shared product code uses
packages/platformcapabilities and ports. - Settings must represent real platform support: local hides cloud usage and
PR-driven auto-archive, and omits machine selection when
remoteMachinesis absent. Gate entries and their background work through capabilities rather than build-kind or environment checks. - Shared packages stay platform-neutral. The public Electron composition
selects
localexplicitly; private Web/mobile entries and cloud composition roots may injectcloudwithout forking those shared packages. - The code-review-viewer build accepts
LODY_RELEASE_VERSIONfor downstream immutable packaging; without it, the public package version is authoritative. - The OSS desktop entry is local-only and must not make authenticated product-cloud requests; public managed-runtime artifact downloads are the explicit exception.
- An absent platform selector resolves to
local; public build scripts must not accept or discover staging/production deployment presets. - Local CLI, renderer, and Electron-main telemetry is hard-disabled even when unrelated PostHog variables exist in the caller's shell.
- Client workflows that require daemon support negotiate integer protocol versions through
MachineMeta.protocolCapabilities; never infer support from the CLI release version. Missing capabilities mean legacy/unsupported. Advertised set and version checks share one binding inpackages/shared/src/machine-protocol-capabilities.tsso a key never travels without its version. - Managed runtime downloads default to the public R2-backed channel owned by
packages/platform/src/runtime-artifacts.ts; local and cloud assembly must use that same constant.LODY_RUNTIME_BASE_URLis only an explicit mirror override. packages/acp-extension-kimiis an isolated submodule workspace. Do not add it to the root pnpm dependency graph; Lody consumes only its separately built, checksummed managed-runtime artifact and versioned ACP extension contract.packages/acp-extension-coreis a public submodule workspace sourced fromLodyAI/acp-extension-core. Keep shared ACP extension contracts there and consume them through the root pnpm workspace; do not duplicate those contracts locally.- Never commit captured user/agent transcripts; fixtures must be synthetic.
- Workspace MCP has exactly two durable layers: catalog entries in the workspace Flock
document and selected ids in each user turn input config. Do not add machine bindings.
Preserve
mcpServerIds: []as an explicit empty selection; dispatch must carry the driving turn's selection into ACP startup rather than rereading session history. - Workspace catalog mutations (MCP servers and Agent Roles) are durable on the local Flock write and shared by an explicit upload that follows it. Settings surfaces resolve on durability and do not wait on or report that upload: the row already exists, the joined room carries the document when a one-shot upload cannot, and a banner about it is something the user can neither act on nor dismiss. What is forbidden is the opposite — reporting a durable write as failed, or rolling one back, because the upload did not go through. The CLI still reports its own sync result to the terminal.
- Agent Roles are one
agentRolerow family in the same workspace Flock document, not a private and a shared catalog: sharing is an ordinary update ofvisibilityon the row. A Role stores no secret — no API key, MCP selection, or memory — andisSensitiveAgentRoleConfigOptionKeyis applied on read as well as on write, because a workspace row reaches every member's client. It DOES pin the permission mode, asrunConfig.modeIdfor legacy ACP modes or the agent's own_permissionoption: permission is a run-config value the agent publishes, not a secret, and a Role that left it out would not be the whole configuration it claims to be. So the composer drops its separate permission button while such a Role is selected. A Role may therefore pin a warning-tone mode (full access / skip permissions), which every surface that hides the permission control must keep visibly marked; what stays out of scope is a Role-level auto-approval POLICY. Settings and mention discovery usecanReadAgentRole/canManageAgentRole; MCP creation resolves an explicit Role id from the workspace catalog without requiring a mention-scoped authorization record. - A Role never falls back.
machineId + agentConfigIdbind the execution site exactly; when the machine, config, or a stored model/mode is unavailable the Role stays listed with the precise reason and stops being mentionable. MCP creation resolves the current workspace catalog row byagentRoleIdbefore Operation acceptance; the canonical Prompt, target, Role revision, and dispatch config are frozen into the accepted Operation so a later edit or delete cannot change its recovery or retry.SessionMeta.agentRoleId/agentRoleRevisionrecord where a Session came from and are display-only.
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.
- today Changed · -8 lines · -57 tokens per session 5ccbcc4acdc9
- 4d ago First seen · 139 lines · 1,846 tokens per session scan A df0fbfb0f7e9
Lody AGENTS.md is an instructions file published in the GitHub repository LodyAI/Lody (869 stars, last pushed today), licensed Apache-2.0. It adds 1,789 tokens to every session, about $0.0089 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
sprawling AGENTS.md
Instructions for 2youg1/sprawling, covering agents.md — how work is done in this repository, the loop, read before you write, one change, five steps and the rules a machine holds.
open-tag CLAUDE.md
Instructions for fancyboi999/open-tag, a project described as: Open-source, self-hostable alternative to Claude Tag — a Slack-style workspace where your team and its AI agents (Claude Code, Codex, GitHub Copilot, and more) work as teammates in channels, threads, DMs, and shared tasks. Your data stays on your machines.
sprawling CLAUDE.md
Instructions for 2youg1/sprawling: See AGENTS.md. It is the whole instruction set for this repository — the loop, what to read first, the five steps of one change, the rules a machine holds, and the language and commit conventions.
open-tag AGENTS.md
Instructions for fancyboi999/open-tag, covering open-tag — guide for ai coding agents, this file is a map, not a manual, conventions, parallel development (worktrees) and isolated dev e2e (on demand).
Octop AGENTS.md
AGENTS.md instructions for TencentCloud/Octop, covering agents.md, 1. collaboration principles, think before writing, simplicity first and surgical edits.
gini-agent AGENTS.md
AGENTS.md instructions for Open-Curiosity/gini-agent, covering gini agent instructions, shape, adrs, boundaries and branches.