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/tjakoen/bread/claude-mdgit clone --depth 1 https://github.com/tjakoen/breadWhat 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 | $0.02342 | $0.02342 |
| Opus 5 | $0.01171 | $0.01171 |
| Sonnet 5 | $0.00468 | $0.00468 |
| Haiku 4.5 | $0.00234 | $0.00234 |
Grade A, and why
bread 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 3d 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 — 141 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md — start here
Onboarding + operating rules for any AI (or human) joining this repo. Read this first, then the docs it points to. Keep it accurate: if you change how the project works, update this file too.
What this is
A no-build, server-rendered hypermedia stack and the things built on it. One direction of dependency — each layer builds only on the layers below it:
batch/ BATCH — the substrate (Bun · Addressable · TypeScript · CSS · htmx); no build step
└─ grain/ GRAIN — an AI-interaction design system + its default theme (the look) + the catalog
├─ MILL/ the markdown CMS (LIVE — renders /notes + the layer docs; its OWN reusable project)
├─ proof/ PROOF — the AI plan board, a mountable layer (plans-as-markdown → kanban)
├─ pantry/ PANTRY — the installable dev-docs + AI cockpit app (`bunx pantry`)
├─ tjakoen.github.io/ THE personal-site app + the composition root
└─ project/ the product — a personal AI assistant; **PAUSED**, docs-only archive
The composition root folded into tjakoen.github.io/ (2026-07-05): the portfolio is now THE app —
it wires batch + grain + mill and runs the site; the hero desk is the reference surface where you
watch the AI act. The portfolio uses MILL for its markdown content; MILL does not build it.
Dependency purity: grain imports nothing from batch except the OpChannel port; MILL depends
on both, never the reverse — a new layer above both, not an extension of either.
The defining idea: a UI where every surface is addressable and operable by both a human
and an AI through one shared vocabulary, with the AI's presence shown as a visible signal
(grain = AI). A human click and an AI decision become the same Intent, flow through
one door (POST /intent → grain/ai/interaction-layer.ts), and return as RenderOps
pushed over SSE. No privileged AI→DOM back channel.
Fuller detail on each layer, and the paused-project history, is in DOCS.md — the
full doc map.
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.
- 3d ago First seen · 141 lines · 2,342 tokens per session scan A 2fa3e443ee4f
bread CLAUDE.md is an instructions file published in the GitHub repository tjakoen/bread (181 stars, last pushed 14d ago), licensed Apache-2.0. It adds 2,342 tokens to every session, about $0.0117 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
webjs AGENTS.md
Instructions for webjsdev/webjs, covering agents.md for webjs, ai-driven development: guardrails for all agents, before starting any work: verify and sync the branch, one task per git worktree, always and skills are routed deterministically, never skipped.
webjs copilot-instructions.md
Instructions for webjsdev/webjs, covering github copilot instructions: webjs app, persistence + scaffold rules (non-negotiable), before starting any work, autonomous mode (sandbox / no-prompt) and mandatory workflow (never skip).
webjs CLAUDE.md
Instructions for webjsdev/webjs, a project described as: WebJs is an AI-first full-stack JavaScript framework built on web components. Any AI model reads the whole stack. No training data required.
fff AGENTS.md
AGENTS.md instructions for dmtrKovalenko/fff, covering to clankers, development commands, building, testing and development tools and code quality.
bunqueue CLAUDE.md
Instructions for egeominotti/bunqueue, covering bunqueue, test isolation rules, architecture flow, directory structure and code guidelines.
gini-agent AGENTS.md
AGENTS.md instructions for Open-Curiosity/gini-agent, covering gini agent instructions, shape, adrs, boundaries and branches.