Borrowing it
Nothing to install: this file belongs to wsauret/flywheel. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/wsauret/flywheel/main/AGENTS.mdgit clone --depth 1 https://github.com/wsauret/flywheelWrote 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/wsauret/flywheel/agents-md)<a href="https://agentmods.dev/instructions/wsauret/flywheel/agents-md"><img src="https://agentmods.dev/badge/instructions/wsauret/flywheel/agents-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/wsauret/flywheel/agents-md"><img src="https://agentmods.dev/badge/instructions/wsauret/flywheel/agents-md.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.08685 | $0.08685 |
| Opus 5.5 | $0.03474 | $0.03474 |
| Sonnet 5.5 | $0.01737 | $0.01737 |
| Haiku 4.5 | $0.00869 | $0.00869 |
Grade A, and why
flywheel AGENTS.md scanned grade A with 1 finding 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.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
| A direct raw external-process launch or shell wrapper in production code | Ordinary launches use `spawnOwned`, `runOwnedCommand`, or `openExternalUrl`; explicitly authorized survival uses only `src/infra/persistent-sub 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.
<coding_guidelines>
Flywheel CLI - Agent Instructions
Runtime: Bun (not Node) | TUI: OpenTUI + SolidJS | Config: TOML
SolidJS requires the "browser" export condition. bin/flywheel handles this via bun --conditions=browser.
This file is the complete guide for contributors. The maintainer also checks out a small private companion repository at the gitignored private/ folder of the clone. Its @private/AGENTS.maintainer.md adds the rules for the maintainer's shared working tree, publication, and release readiness. That import is optional: a contributor's checkout has no private/ folder, the import then loads nothing, and every rule a contributor needs is in this file.
1. Hard Rules
-
MAXIMIZE ELEGANCE, NOT MINIMIZE CHURN! THIS IS THE MOST IMPORTANT PRINCIPLE FOR EVERY DESIGN! Always pick the design that maximizes elegance no matter how big the refactor. Better now than months down the line when there are even more dependencies in place.
-
docs/elegance.mdis Flywheel's canonical project-specific definition of elegance and the rubric for architecture audits. Read it in full when judging whether the project is maximally elegant. This file operationalizes much of that standard for coding agents, but not in full. -
Handle only reachable states. A state is reachable when a person or an external producer can cause it through a supported entry point: a client surface, the CLI, the published protocol as shipped clients use it, or an external boundary such as a provider stream, a subprocess, or the file system. A state that only a direct internal call or a test fixture can create is hypothetical: add no code for it, and do not treat a test that asserts it as a product requirement. Before you call a failure a regression, trace how a person reaches it. When a test fails on a state that no person can reach, rewrite the test around the reachable flow instead of changing the product.
-
This repository opts into the living product-intent authority indexed by
docs/product-intent/README.md. Before changing or judging user-observable Flywheel behavior, read the index and the relevant domain file. A conflict with the persisted corpus is a bug or blocker, unless the person states a different intent; then the corpus changes to match it. -
Before judging whether a subscription route follows its provider's terms, read
docs/provider-terms.md. This covers the Claude engine route, the sign-in client IDs, the request headers that identify a client, and the use of several accounts of one provider. Flywheel follows precedent: a practice that widely used open-source agents ship publicly is acceptable. The page's section "How a review uses this page" lists the practices that the maintainer settled on that basis. A review or an audit never reports one of them as a finding, a blocker, or a risk, also when it finds provider terms that the page does not quote. Only the reopen events that the page names reopen a question. -
Repository code, comments, and documents name no other coding-agent harness.
docs/provider-terms.mdis the one place that names such projects as precedent. Credit for borrowed work goes inTHIRD-PARTY-NOTICES.mdand the license bundles, and a file or directory that a notice cites by path keeps its upstream references. The only other names allowed are external identifiers that Flywheel must spell exactly to interoperate, such as an environment variable, a header, or a model id, and benchmark arms that run another harness as the comparison treatment.README.mdand the website home pagewebsite/src/pages/index.astromay name other harnesses, because they market Flywheel to their users. -
This repository opts into the living architecture-intent authority indexed by
docs/architecture-intent/README.md. Before changing or judging Flywheel's internal organization, read the index and the relevant domain file. A conflict with the persisted corpus is a bug or blocker, unless the person states a different intent; then the corpus changes to match it.docs/terminology.mdis the lexicon for Flywheel's terms.
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 Changed b99b81da1c9d
- 3d ago Changed · +32 tokens per session a202d0f1c4bb
- 4d ago First seen · 274 lines · 8,653 tokens per session scan A e53b33615d2c
flywheel AGENTS.md is an instructions file published in the GitHub repository wsauret/flywheel (19 stars, last pushed yesterday), licensed Apache-2.0. It adds 8,685 tokens to every session, about $0.0347 per session on Opus 5.5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-11-03.
flywheel AGENTS.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| flywheel AGENTS.mdwsauret/flywheel | Instructions file | 19 | yesterday | Apache-2.0 | A | 8,685 | Codex, OpenCode |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 143k | yesterday | MIT | A | 7,461 | Codex, OpenCode |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | yesterday | MIT | A | 4,680 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 5,516 | GitHub Copilot |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 3,093 | GitHub Copilot |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 274k | 2d ago | MIT | A | 4,313 | Codex, OpenCode |