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/davisjam/model-based-agentic-software-engineering/claude-mdgit clone --depth 1 https://github.com/davisjam/model-based-agentic-software-engineeringWhat 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 | $0.04237 | $0.04237 |
| Opus 5 | $0.02119 | $0.02119 |
| Sonnet 5 | $0.00847 | $0.00847 |
| Haiku 4.5 | $0.00424 | $0.00424 |
Grade A, and why
model-based-agentic-software-engineering 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 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 — 225 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md — agent-governance-mechanisms
This repo is a pattern catalogue: 85 governance mechanisms that keep a fleet of autonomous coding agents productive while bounding their failures, each written as a Gang-of-Four-style design pattern. It is published as a static GitHub Pages site and embedded in a parent repo as a git submodule.
catalog.py is the one tool — stdlib-only, no dependencies (so python3 catalog.py … runs on a
fresh checkout with nothing installed). Do not add pip dependencies; keep it clone-and-run.
Site is a preview; the book claims and expands
The published site — the landing, the catalogue views, the entry pages — is a preview: concise
framings and entry points. The book (under book/) is where each idea is claimed and expanded.
Two jobs, one for each surface.
The standing implication: book coverage ⊇ site framings. Any conceptual framing that appears on the
site should have a fuller treatment in the book. When you add a framing to the site's landing prose (the
LANDING_INTRO, the schools, the ways, the spectrum axis) or an entry page, either the book already
expands it or you owe it a home there. Site-only material is limited to adoption and navigation —
DIY-vs-install, the nav cards, the quick-start; those need no book counterpart. Everything conceptual does.
What is a mechanism?
A mechanism is a recurring, failure-killing pattern of governance — written like a Gang-of-Four design pattern: the failure class it prevents, the shape that prevents it, and why it is not just the cheaper thing everyone already does. A mechanism earns its own entry only when it clears three bars:
- It kills a failure class, not a one-off bug. If you can't name the recurring failure, it isn't a mechanism yet.
- It's distinct. It does something the naive or adjacent alternative can't — every entry must survive its own "Why it's not just [X]" section. Two entries built on the same pattern are distinct only when they vary a named axis (lock cardinality, object model, enforcement scope, domain); if nothing varies, merge them.
- Its examples instantiate it. Every example — in Motivation, in "Why it's not just…", in Known uses — must be a real case of this mechanism, not a sibling's. (A relational config-field ⊆ sample check is a coherence lint, not a per-source semantic lint — so cite it under coherence only.) An example borrowed from a neighbour blurs the exact boundary the catalogue exists to draw.
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 · 225 lines · 4,237 tokens per session scan A af7a638d70b3
model-based-agentic-software-engineering CLAUDE.md is an instructions file published in the GitHub repository davisjam/model-based-agentic-software-engineering (18 stars, last pushed 2d ago), licensed MIT. It adds 4,237 tokens to every session, about $0.0212 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
agent-ready-repo AGENTS.md
Instructions for eugenelim/agent-ready-repo, covering agents.md, project overview, rule lookups, documentation and development workflow.
Forge-Agentic-Coding-CLI CLAUDE.md
Instructions for hoangsonww/Forge-Agentic-Coding-CLI, covering claude.md, identity of this repo, build & verify commands (canonical), repository map and conventions.
claude-conductor CLAUDE.md
Instructions for superbasicstudio/claude-conductor, covering claude.md, critical context (read first), project architecture, core component and planned documentation modules.
cdv copilot-instructions.md
Instructions for azank1/cdv, covering cdv — agent instructions, required on every message and memory & recovery (v0.8).
vulyk CLAUDE.md
Instructions for Black-coffe/vulyk, covering vulyk constitution, the five laws, working with a frontier model, complexity routing (decide before working) and token economy (non-negotiable).
myace CLAUDE.md
Claude Code instructions for niels-emmer/myace, a project described as: MyACE makes your AI coding agent's rules, skills, and workflows portable.