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/agents-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 | $0.02202 | $0.02202 |
| Opus 5 | $0.01101 | $0.01101 |
| Sonnet 5 | $0.00440 | $0.00440 |
| Haiku 4.5 | $0.00220 | $0.00220 |
Grade A, and why
model-based-agentic-software-engineering 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 — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — running editorial folds on the MAGE book with a fleet
This file is the operational playbook for making large editorial changes to the book
(book/) with dispatched agents. It extends CLAUDE.md §"Working on this repo with a fleet"
(single-live-writer + parallel drafting), which you should read first. Where CLAUDE.md states
the rules, this file gives the worked process — the one that lands a multi-part copyedit pass
in one sitting without serializing on main and without a merge that fights 100 generated files.
The method has four patterns. Use them together: draft in parallel → fold in parallel worktrees → merge by source-checkout → resolve the model layer once at the end.
0. The one fact everything follows from
main has a single-live-writer rule for exactly one reason: the pre-commit hook rebuilds the
site and force-stages the regenerated .html + derived book-models/* on every commit, so two
agents committing to the same branch collide on those generated files. Everything else about the
book is deterministic from source. That single fact is what makes the patterns below safe:
- Two folds on separate worktree branches never collide (separate git indexes).
- Two folds touching disjoint source trees (different Parts / the appendices) never conflict on content — only on regenerated files, which you never merge, you rebuild.
- A PDF renders from the working tree, not from
HEAD— so you can build a review PDF from staged-but-uncommitted source.
1. Pattern — disjoint-worktree parallel folds
When several folds touch disjoint source trees (e.g. Part II prose, Part III prose, appendices),
run them concurrently in separate git worktrees instead of serially on main.
Setup, per fold:
git worktree add -b <fold>-fold <path> HEAD(branch from currentmain; put<path>in a scratch dir).- ★ Copy the drafts into the worktree.
book/_design/drafts/is gitignored, so a fresh worktree checkout does not contain the EDIT-MAP or the authored figures. Without this the agent has nothing to fold:cp -r book/_design/drafts/<effort>-<date> <worktree>/book/_design/drafts/(they stay untracked in the worktree — the hook never stages them). - Dispatch the fold agent with its CWD = the worktree and a brief that says: you are on branch
<fold>-fold, confirm withgit branch --show-current, do NOT touchmain, other folds run concurrently on disjoint trees — ignore them.
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 · 144 lines · 2,202 tokens per session scan A dafbd893dd27
model-based-agentic-software-engineering AGENTS.md is an instructions file published in the GitHub repository davisjam/model-based-agentic-software-engineering (18 stars, last pushed yesterday), licensed MIT. It adds 2,202 tokens to every session, about $0.0110 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
zola-mcp CLAUDE.md
Instructions for chrischall/zola-mcp, covering zola-mcp, commands, architecture, environment and auth resolution (four-path).
rubydex AGENTS.md
AGENTS.md instructions for Shopify/rubydex, covering instructions, project overview, documentation, ruby gem and structure.
AmbyKit CLAUDE.md
Instructions for ambystechcom/AmbyKit, covering claude code — notes for the ambykit repo and claude-specific.
lumen AGENTS.md
AGENTS.md instructions for ory/lumen, covering agents.md, repository expectations and commands.
voicetree CLAUDE.md
Claude Code instructions for voicetreelab/voicetree: ⛔ This is a NORMAL, WRITABLE checkout of THIS machine's own branch $VTDEVBRANCH (machine-local, set in /.env — e.g. dev-mac / dev-remote; never a literal in-repo; safe default dev-new). Edit & commit directly here (worktrees optional: vt-worktree ). vt-sync to pull…
ai-auto-work CLAUDE.md
Claude Code instructions for chaohong-ai/ai-auto-work, covering gamemaker - ai native 游戏创作平台, 项目简介, ai 工作目录约定 and 按需加载（不默认装载）.