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/blake-simpson/belmont/agents-mdgit clone --depth 1 https://github.com/blake-simpson/belmontWrote 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/blake-simpson/belmont/agents-md)<a href="https://agentmods.dev/instructions/blake-simpson/belmont/agents-md"><img src="https://agentmods.dev/badge/instructions/blake-simpson/belmont/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 | $0.12732 | $0.12732 |
| Opus 5 | $0.06366 | $0.06366 |
| Sonnet 5 | $0.02546 | $0.02546 |
| Haiku 4.5 | $0.01273 | $0.01273 |
Grade C, and why
belmont AGENTS.md scanned grade C with 2 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.
Tells the agent never to refusehighAnti-refusal
Suppressing the ability to decline removes a core safety control; a later harmful request then succeeds.
because upgrading Belmont must never refuse a run that worked the day before Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- `install.sh` (root) — Public curl-pipe-sh installer for end users. How it starts
The opening of the file, as written. The whole thing — 346 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS
This file provides guidance to Ai Agents when working with code in this repository.
Self-learning loop — consult knowledge/KNOWLEDGE.md first
Belmont's architectural memory lives in the knowledge/ tree at the repo root. knowledge/KNOWLEDGE.md is a thin index that routes you to self-contained topic entries — one per concept — organised by domain. The index is cheap to read in full; entries are read on-demand based on what you're about to touch.
Rules:
- Before designing a fix for any bug in the auto loop, skills, worktree lifecycle, merge strategy, verification, agent dispatch, or state model: read
knowledge/KNOWLEDGE.md. Match the "Read when you're about to…" column to your task; open only the entries that match. Most sessions need 1–3 entries, not the whole tree. - Every entry has a stable skeleton (
Why this matters→Invariant→How it's enforced→Failure mode if you break it→Don't re-do→Evidence→Revisions). Jump toDon't re-dobefore proposing a "why don't we just…" approach — a lot of them are already ruled out with reasoning. - Cross-domain entries live in
knowledge/cross-cutting/and carry aDomains:header line you can grep. Single-domain entries live inknowledge/<domain>/. - Amend, don't append. When you produce a durable insight (a new invariant, a design option deliberately rejected, a rough edge discovered), edit the relevant entry in place and add one line to its
Revisionsfooter:YYYY-MM-DD — what changed. Do not add dated blocks at the top. - New concept → new entry. If a session uncovers a genuinely new topic that doesn't fit any existing entry, create
knowledge/<domain>/<topic>.mdfollowing the skeleton, add a row toKNOWLEDGE.md's routing table. Do not inline new concepts into existing entries to avoid the file count going up — file count going up is fine when the concept is real. - Cross-topic chronology is
git log -- knowledge/. There is no global decision log; it would duplicate git and bloat unbounded. - Treat
AGENTS.mditself as live: if a change in a knowledge entry implies agents should approach work differently on first read, update the relevant section ofAGENTS.mdso the guidance is immediate. KeepAGENTS.mdterse — prefer linking into the knowledge tree over duplicating detail. - When you update
AGENTS.md, say so in your end-of-turn summary and note which entry motivated the change.
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 · 346 lines · 12,732 tokens per session scan C 5e630ec131fb
belmont AGENTS.md is an instructions file published in the GitHub repository blake-simpson/belmont (4 stars, last pushed 15d ago), licensed Apache-2.0. It adds 12,732 tokens to every session, about $0.0637 per session on Opus 5. A static security scan graded it C with 2 findings (tells the agent never to refuse, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
deer-flow copilot-instructions.md
Copilot instructions for bytedance/deer-flow, covering copilot onboarding instructions for deerflow, 1) repository summary, 2) runtime and toolchain requirements, 3) build/test/lint/run - verified command sequences and a. bootstrap and install.
turbo-flow CLAUDE.md
Claude Code instructions for marcuspat/turbo-flow, covering claude.md — turbo flow, what this repo is, key files and conventions.
work-buddy CLAUDE.md
Claude Code instructions for KadenMc/work-buddy, covering work buddy, mcp gateway, capability vs workflow, workflow consent (composable) and session init (mandatory).
orc AGENTS.md
AGENTS.md instructions for spencermarx/orc, covering openspec instructions, landing the plane (session completion) and open code review instructions.
abcd CLAUDE.md
Instructions for intentdriven/abcd, a project described as: abcd: An experimental configuration layer for intent-driven development.
abcd GEMINI.md
Instructions for intentdriven/abcd, a project described as: abcd: An experimental configuration layer for intent-driven development.