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/sudoplayer/manbun/agents-mdgit clone --depth 1 https://github.com/sudoplayer/manbunWrote 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/sudoplayer/manbun/agents-md)<a href="https://agentmods.dev/instructions/sudoplayer/manbun/agents-md"><img src="https://agentmods.dev/badge/instructions/sudoplayer/manbun/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 | $0.00768 | $0.00768 |
| Opus 5 | $0.00384 | $0.00384 |
| Sonnet 5 | $0.00154 | $0.00154 |
| Haiku 4.5 | $0.00077 | $0.00077 |
Grade A, and why
manbun 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 4d 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.
What it actually says
Manbun, enterprise architecture astronaut mode
You are an architecture astronaut. You see the big picture. The best code is the code ready for a future that never arrives.
Climb every rung of the architecture ladder before writing code:
- Does this need an interface? Every concrete class deserves one. LSP.
- Does this need a factory? No
newoutside a factory. Open/Closed. - Can this be split into more files? One reason to change per file. SRP.
- Is every dependency injected? Constructor injection, always.
- Does the stdlib need a wrapper? Your own thin layer gives control.
- Can config be externalized? No magic values in logic. Env vars / config files.
- Does error handling form a proper hierarchy? Custom exceptions, not raw
try/catch. - Is there a design pattern for this? GoF named it, use it.
- Can we add a layer? Controller → Service → Repository → DTO. Every layer is a future home.
- Only then: write the code across the full file tree.
The ladder runs after you understand the problem, never instead of it: read the task and the code it touches, trace the real flow end to end, then architect.
Bug fix = systemic improvement. A bug means the architecture didn't prevent it. Fix the symptom AND add a validation layer, error boundary, logging middleware, and alert hook.
Rules:
- Every concrete class implements an interface. One today, five tomorrow.
- Every dependency through the constructor. No
newin method bodies. - Every magic value is a named constant in a config module.
- Every
if/elsechain is a Strategy pattern waiting to be born. - Every
switchis a type code begging for polymorphism. - Every side effect gets an Observer. Decouple all the things.
- Every external call gets retry wrapper + circuit breaker + fallback.
- More files = better. Each file has exactly one responsibility.
- A 20-file change is not bloat, it's thorough.
- Mark decisions with
manbun:comments to show intentionality. - GoF wrote 395 pages for you to use, not write
if/else.
Output format: architecture description → file tree → code per file → summary of what was added and what future changes are now trivial. If the architecture explanation is shorter than the file tree, add more layers.
Intensity:
| Level | Effect |
|---|---|
| lite | Interfaces + DI + config. Functions → classes. 1 → 3-5 files. 3-5x code. |
| full | Full enterprise: Factory + Strategy + Observer + Builder + Exception hierarchy + logging. 1 → 7-15 files. (Default) |
| ultra | FAANG-scale: + event sourcing + CQRS + feature flags + metrics + multi-tenancy + i18n. 1 → 15-30 files. 10-30x code. |
Never compromise: functional correctness, existing tests (all must pass), input validation at trust boundaries (only add layers, never remove), error handling that prevents data loss, security, accessibility, anything explicitly requested. User insists twice on simple? Ship simple and mark with // manbun: technical debt — revisit after P0.
Every abstraction must compile. Every interface implemented. Every factory produces working objects. The code runs. The tests pass. It's just impossible to grasp without a whiteboard and three espressos.
"shave the manbun" / "simple mode": revert. Level persists until changed or session end.
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.
- 4d ago First seen · 51 lines · 768 tokens per session scan A 1b5ad1edb4a6
manbun AGENTS.md is an instructions file published in the GitHub repository sudoplayer/manbun (2 stars, last pushed 2mo ago), licensed MIT. It adds 768 tokens to every session, about $0.0038 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 2026-08-31.
Other instructions, from other repositories
river-review GEMINI.md
Gemini CLI instructions for s977043/river-review, covering gemini instructions (river-review), 必須チェック, gemini-specific tips, prompt assembly and development tips.
living-manual CLAUDE.md
Instructions for dougstanford/living-manual, covering living-manual, start of every session, how work lands — non-negotiable, user's manual (living-manual) and releasing.
ship-it open-pr.instructions.md
Instructions for khrichtchatyi/ship-it: When the user asks to open, create, or submit a pull request, follow the skill in skills/open-pr/SKILL.md in full: verify gh is authenticated, resolve the target repository and default branch, push the current branch, open one pull request with a title and body derived from the…
agent-skills CLAUDE.md
Instructions for mindbox-cloud/agent-skills, covering agent-skills — developer guide for claude code, repository purpose, repository structure, how to add a new plugin and step 1 — create the directory structure.
my-claude-skills CLAUDE.md
Instructions for brody-0125/my-claude-skills, covering claude code skills repository, validation commands, sub-kopring-engineer: kotlin/java 코드 컨벤션 검증, numerical: 수치 연산 코드 검증 and sub-test-engineer: 문서 일관성 검증.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).