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/max-listov/stitchkit/agents-mdgit clone --depth 1 https://github.com/max-listov/stitchkitWhat 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 | $0.04943 | $0.04943 |
| Opus 5 | $0.02472 | $0.02472 |
| Sonnet 5 | $0.00989 | $0.00989 |
| Haiku 4.5 | $0.00494 | $0.00494 |
Grade A, and why
stitchkit 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 2d 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 — 338 lines — stays where its author put it; the contents beside it link to each section on GitHub.
stitchkit — agent guide
Contract-first backend framework for Bun and Node. One defineContract() → an
HTTP API, MCP tools, AI-agent tools and a typed client.
Two roads — pick yours
📦 Building an app with stitchkit? This file is not for you — it is about changing the framework. Start at the README quick start and the user guide. In your own project your agent's entry point is
node_modules/stitchkit/llms.txt(ships in the package); Claude Code users can also drop the repo'sskills/stitchkitinto.claude/skills/. They map the whole consumer surface.🔧 Developing stitchkit itself? You're in the right place. This file is the canonical, tool-agnostic guide — the rules, architecture, breaking-change and release flow. The hands-on contributor workflow (setup, commands, git hooks, local development against a consuming app, PRs) is in
CONTRIBUTING.md. Design rationale per rule is an ADR indocs/decisions/; tasks live indocs/backlog/.
Rules
- NEVER ship a competing WebSocket or hook engine — wrap Socket.IO
(
createSocketIOClient/createSocketIOServer) andreact-query-kit(createCursorQuery). → ADR 0008 - NEVER write
ascasts in business logic. A cast that ships is a boundary: the loose↔typed bridges ininternal/typed.ts, adapters over untyped external emitters (Socket.IO, the event bus, the cache bridge), and the generic bridges inbrowser/client.tswhere a scoped client surface is rebuilt from a wider one. Each must carry a comment saying why. A new cast anywhere else means the types are broken upstream; fix them there. Nothing enforces this mechanically — it is a review rule over a small, countable set (thirteen inpackages/core/src), not a gate. → ADR 0003 - ALWAYS keep the core Web Fetch-clean —
createHandlertakesHandlerConfig(no Bun types). Bun APIs live only increateServerandstitchkit/server. → ADR 0013 - ALWAYS Zod-first — a schema is the source of truth, types come from
z.infer. Never hand-write a duplicate type. - NEVER make the project declaration a condition — of a build, a test, a
start path or a check. A project with no
project.jsonis a complete project;stitchkit/declarationis a leaf nothing else in the core imports, andpackages/core/tests/project-declaration.test.tskeeps it one. A repository only one tool can bring up is a fork, not a dependency. → ADR 0104 - ALWAYS keep the core generic — no domain model. Scopes are free strings,
there is no billing,
sourceis transport-only. → ADR 0002 - Transport and hooks use
RuntimeContext(loose); handlers useHandlerContext(typed). Do not cast between them. → ADR 0003 - A new architectural decision → a new ADR in
docs/decisions/and a row indocs/decisions/README.md(keep the index in sync). A new idea → a file indocs/backlog/inbox/. Seedocs/README.md. - A completed backlog item may claim test coverage only by naming the exact
test file and test case in its
Что сделаноsection. - ALWAYS run
bun scripts/verify.ts --releasebefore a release commit, and let thepre-pushhook reuse that exact-tree result — see What runs where below.verifyis the whole portable local gate and it runs every portable gate CI runs: lint, typecheck, tests, the Postgres agent-store lane, build, the Next-SSR and Node smokes, the packed consumer lane, the packed starter lanes and the supervised PM2 lane. Its prerequisites are listed inCONTRIBUTING.mdand all of them arrive withbun installexcept a reachable PostgreSQL and the Playwright browsers.
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.
- 2d ago First seen · 338 lines · 4,943 tokens per session scan A fd5ec8ebdae8
stitchkit AGENTS.md is an instructions file published in the GitHub repository max-listov/stitchkit (4 stars, last pushed 2d ago), licensed MIT. It adds 4,943 tokens to every session, about $0.0247 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
keryx CLAUDE.md
Instructions for actionhero/keryx, covering claude.md, project overview, monorepo structure, development environment and environment setup.
wa-automate-nodejs AGENTS.md
AGENTS.md instructions for open-wa/wa-automate-nodejs, covering repository instructions, commit policy, no ai attribution, commit grouping and gitmoji reference.
wa-automate-nodejs copilot-instructions.md
Copilot instructions for open-wa/wa-automate-nodejs, covering switchboard configuration for github, available mcp tools, messaging (cross-ide), workflow management and terminal management.
igniter-js GEMINI.md
Gemini CLI instructions for felipebarcelospro/igniter-js, covering 1. identity and profile, 2. about the igniter.js monorepo, 3. personality and communication, 4. lia's core responsibilities (the 4 pillars) and 5. technical guidelines and methodology.
igniter-js AGENTS.md
AGENTS.md instructions for felipebarcelospro/igniter-js, covering lia - ai agent for igniter.js, 1. identity & mission, core mission, key responsibilities and 2. project overview.
igniter-js writing-style.instructions.md
Instructions for felipebarcelospro/igniter-js, covering ✍️ unified documentation style guide (for llms & authors), 🎯 core style principles (applies to all documentation), 📘 documentation (docs): writing style, installation and quick start.