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 skills/baixinghao/intent-gate/contract-codingnpx skills add baixinghao/intent-gate --skill contract-codinggit clone --depth 1 https://github.com/baixinghao/intent-gateWhat 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 | $0.00121 | $0.01117 |
| Opus 5 | $0.00060 | $0.00558 |
| Sonnet 5 | $0.00024 | $0.00223 |
| Haiku 4.5 | $0.00012 | $0.00112 |
Grade A, and why
contract-coding 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.
What it actually says
Contract-Driven Coding(附加层,非编码 skill)
本 skill 是附加层,不是编码 skill。 你的项目自有编码 skill、superpowers、 CLAUDE.md/AGENTS.md 规约继续拥有"怎么写代码"(架构、风格、模式)的全部决定权。 本 skill 只管一件事:当 intent-gate 契约存在时,代码从契约生成,且绝不违背契约。 代码风格/结构上的任何冲突,以项目自有规则为准;契约保真这件事,本 skill 不让步。
契约路由(多需求并存,🔴 禁止猜)
- 人类点名了需求 → 精确匹配
.harness/requests/{需求名}/。没有同名目录 = 该需求没有契约,走第 3 条——🔴 禁止模糊匹配到别需求的契约 ("授信提交"套"提现确认"的 mermaid 是事故,不是便捷)。文件夹名即唯一标识, 文件系统保证不重名,无需另造 ID。 - 人类没点名 → 列出
.harness/requests/下的契约清单让人类选,禁止自己挑。 - 点名的需求无契约、但项目里已有其他契约(说明本工作流已启用)→
告知人类"该需求还没有契约",给两条路:先跑
requirement-alignment出契约, 或人类明确说"不要契约直接写"——后者合法,人类指令优先。 (仅当整个项目没有任何.harness/requests/时才完全静默、按原方式编码。) - 采用契约前交叉验证:任务的领域词汇(实体/动作)应命中契约的术语表; 明显对不上 → 停,向人类确认是否拿错了契约。
开工门禁(机械项,无商量)
- 路由确认契约后读 frontmatter:
status: approved才动工;pending_review/blocked→ 停下报回人类。approved 只有两条合法来源: 蓝军 PASS 或人类直接拍板——红蓝评审是可选项,人类说"不用评审,我拍板" 即合法开工,任何人不得强制走红蓝。 - 读
_review/lint-report.md:CRITICAL > 0 → 契约自身带机械错误, 禁止对着它编码,报回人类先修契约。
🔴 门禁拒绝的正确动作是停下报回人类——契约状态是人类拍板的事,
不是"再分析一遍"的事。禁止因为 status ≠ approved 或 lint 未归零就转入
requirement-alignment 重启/续跑意图对齐流程。
从契约施工
- mermaid 状态机 / 时序图 / 决策表 + DDL 就是规格。每个实现任务可回溯到锚点: 哪条边 → 哪个接口/方法;哪条 BR 规则 → 哪个校验分支;哪张表 → 哪个实体/Mapper。
- 只实现契约里有的。 图里没有的逻辑不许出现在代码里——无契约逻辑就是 意图对齐要消灭的"静默猜测"在编码期的复活。
- 术语沿用契约词汇表(wiki 对齐版),禁止新造命名。
漂移 = 停线
编码中发现契约错了/不够用(现实逼着偏离图):
- 停下,禁止顺手绕过去;
- 走
dispatch_question升级(对话框通道直接向人类提问); - 先改契约(重新对齐 + 重跑 lint),再改代码。
mermaid 与代码漂移即 bug——无论哪个方向。
完工
- 契约若有修订,重跑
lint_summary(CRITICAL 归零)。 - 说明每处改动实现了契约的哪些锚点,让评审能从代码回溯到契约。
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 · 62 lines · 0 tokens per session scan A f9712fbf8c5d
contract-coding is a skill published in the GitHub repository baixinghao/intent-gate (2 stars, last pushed 18d ago), licensed MIT. It adds 121 tokens to every session and 1,117 once invoked, about $0.0006 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 skills, from other repositories
build-teaql-app
Build or change a TeaQL application in Java, Rust, Go, Swift, Python, C#/.NET, or TypeScript, including Kotlin/JVM applications that consume Java-generated libraries. Mandatory order: first draft and save a complete KSML model, then verify the client and evaluate that saved model, repair it through repeated evaluation…
mcpfusion-development
How to build production MCP servers with MCP Fusion using the MVA (Model-View-Agent) pattern. Use this skill whenever writing, modifying, or reviewing MCP Fusion code — including tools, Presenters, Models, middleware, prompts, routers, tests, or server configuration. Activate even when the user just says "create a…
vinkius-deploy
How to deploy MCP servers to Vinkius Edge using mcpfusion deploy. Use this skill whenever deploying, configuring remote settings, preparing an entrypoint for edge deployment, or troubleshooting deploy-related errors. Activate when the user says "deploy", "publish to edge", "mcpfusion deploy", "push to Vinkius"…
verifying-compose-ui
Verifying Compose UI by rendering components or previews and inspecting the resulting images and state transitions. Activate for visual behavior that requires runtime rendering; use ordinary tests for nonvisual logic.
raindrop-mcp-publishing
Automates and manages the complete publishing workflow for @adeze/raindrop-mcp from version bumping through npm registry deployment.
mcp-testing
MCP Testing Strategies with Vitest, Inspector, and Integration Tests.