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/arrrrny/zuraffa/agents-mdgit clone --depth 1 https://github.com/arrrrny/zuraffaWrote 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/arrrrny/zuraffa/agents-md)<a href="https://agentmods.dev/instructions/arrrrny/zuraffa/agents-md"><img src="https://agentmods.dev/badge/instructions/arrrrny/zuraffa/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 | $0.02587 | $0.02587 |
| Opus 5 | $0.01293 | $0.01293 |
| Sonnet 5 | $0.00517 | $0.00517 |
| Haiku 4.5 | $0.00259 | $0.00259 |
Grade A, and why
zuraffa 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 — 274 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI Agents Guide for Zuraffa
This file defines the current v5 contract for AI agents working in Zuraffa projects.
Generation contract: use the canonical v5 workflow
When you need to create or evolve architecture code, use this sequence:
zfa entity createzfa makezfa cache adapter <EntityName>(if entity needs cache/local storage)zfa build
Canonical example
zfa entity create -n Product \
--field id:String \
--field name:String \
--field price:double
zfa make Product \
--preset=crud \
--methods=get,getList,create,update,delete \
--with=vpc \
--state \
--di \
--test
zfa cache adapter Product # optional: register Hive adapters if caching needed
zfa build
Hard rules
- Do not use the removed legacy one-shot generator.
- Prefer
zfa makeoverzfa feature.zfa feature scaffoldis only a wrapper over the feature preset. - Do not hand-create entities. Use
zfa entity create. - Do not call
build_runnerdirectly in normal agent flows. Usezfa build. - Do not invent alternate folder structures. Zuraffa v5 assumes a fixed domain root.
STOP-ON-ROADBLOCK RULE (HARD, NON-NEGOTIABLE — HARDCODED)
The FIRST time a zfa command errors, OR succeeds but produces output other than
expected (non-compiling generated code, missing files the generated code references,
missing symbols the generated code uses, silent 0-output builds, crashes), the agent
MUST STOP IMMEDIATELY. This overrides any "keep going" instinct.
- Your first instinct to STOP is the correct one. Trust it. When something is off, stop and report. Do not rationalize it away, do not "just try one more command," do not assume it is a stale cache / wrong flag / your own mistake. Report it as a zuraffa gap first; the maintainer decides if it is real.
- Never continue the generation/build task after a roadblock. Do NOT work around it
with different flags,
--forcere-runs, explicit--methods, re-making only the "id-having" entities, hand-written patches, or "trying the next entity." Bounded or partial progress is STILL a violation — if a gap blocks one entity, it blocks the whole workflow. The smoke-test goal is to build a seamless app only via zfa commands so we can find and fix zuraffa gaps, not to produce a working mock by skirting them. - Never hand-write code to route around a zfa misfire. Every Dart file in
apps/zikzak_demomust come from azfacommand; ifzfacannot produce it, that is the bug to report. - Document in the tracking file (
apps/zikzak_demo/PROGRESS.md) exactly:- the command that was run,
- what was expected,
- what was actually output,
- the root cause (trace it in zuraffa source if possible).
- File a GitHub issue on
arrrrny/zuraffawith that same detail (repro, expected, actual, root cause, suggested fix). - End the goal / stop the run. Wait for the issue to be MERGED (not just opened
or PR'd) before resuming. Resume ONLY when a new goal is invoked that references
apps/zikzak_demo/PROGRESS.md— the resume marker records the exact stopping step so the next run picks up from there.
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 · 274 lines · 2,587 tokens per session scan A b1bde70687c8
zuraffa AGENTS.md is an instructions file published in the GitHub repository arrrrny/zuraffa (5 stars, last pushed today), licensed MIT. It adds 2,587 tokens to every session, about $0.0129 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 2026-09-03.
Other instructions, from other repositories
flutter_pos CLAUDE.md
Instructions for elrizwiraswara/flutter_pos, covering claude.md - flutter project conventions, architecture, detailed structure, data flow and implementation approach.
prprompts-flutter-generator GEMINI.md
Instructions for Kandil7/prprompts-flutter-generator, covering gemini cli extension for prprompts v5.1.0, v4.0: full automation, what is gemini cli?, prerequisites and 1. install node.js.
flutter-intellij CLAUDE.md
Claude Code instructions for flutter/flutter-intellij, covering flutter intellij plugin — claude code guide, additional rules and agent skills & workflows.
wind dynamic.instructions.md
Instructions for fluttersdk/wind, covering dynamic json renderer (lib/src/dynamic/), json schema, security model (whitelist first), state binding (id-keyed widgets) and action dispatch.
wind CLAUDE.md
Claude Code instructions for fluttersdk/wind, covering wind ui, definition of done, post-change sync (five surfaces, every code change), commands and coverage policy (binding).
dart-sdk-skills AGENTS.md
AGENTS.md instructions for RandalSchwartz/dart-sdk-skills, covering ai agent developer handbook (agents.md), 🎯 repository purpose, 🔄 runbook 1: updating for new dart sdk releases, 🔄 runbook 2: updating for new flutter framework releases and 📏 quality & style guidelines.