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/wangliquan/specforge/claude-mdgit clone --depth 1 https://github.com/WangLiquan/specforgeWrote 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/wangliquan/specforge/claude-md)<a href="https://agentmods.dev/instructions/wangliquan/specforge/claude-md"><img src="https://agentmods.dev/badge/instructions/wangliquan/specforge/claude-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 | $0.00918 | $0.00918 |
| Opus 5 | $0.00459 | $0.00459 |
| Sonnet 5 | $0.00184 | $0.00184 |
| Haiku 4.5 | $0.00092 | $0.00092 |
Grade A, and why
specforge CLAUDE.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 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.
How it starts
The opening of the file, as written. The whole thing — 31 lines — stays where its author put it; the contents beside it link to each section on GitHub.
SpecForge
两个 skill：specforge-draft（把模糊需求拷问成可视化 *.spec.html）+ specforge-verify（拿 spec 逐条验落地产物，双出口：把判定回写进自产 spec.html + 在对话里给差距清单，不当 fixer，修复交常规编辑）。定位是「任何要想清楚、定下来、再回头验收的事」，代码只是最常见的产物——不是前提（详见两个 SKILL.md 的 description）。
源 vs 派生产物（红线）
- 源：
lib/（渲染/抽取逻辑）、schema/spec.schema.json - 派生：
skills/*/scripts/*.mjs（由lib/esbuild 打包）、skills/*/references/*.json（由schema/同步）、.claude-plugin/{plugin.json,marketplace.json}的version字段（由package.json.version经scripts/sync-version.mjs同步）。 - 改了
lib//schema/，或 bump 了package.jsonversion 后，必须跑npm run build重新生成派生文件再提交，否则npm run check（CI 同款）的一致性校验会拦。 - 禁止手改
skills/*/scripts/*.mjs或这两处 plugin/marketplace 的version——都是派生产物，下次 build 会覆盖。版本号唯一真源是package.json.version。
关键契约
- spec.html 数据岛 DOM id =
specforge-data（draft 注入、verify 抽取的唯一契约）。改动它须同步：lib/render-spec.mjs注入、lib/extract.mjs正则、showcase.html、test/*断言。 - verify 两条 CLI：
extract-main.mjs抽 spec JSON 给 agent 逐条判定；annotate-main.mjs把 verdicts 合并进 spec 并原地重渲染 spec.html（注入徽标/覆盖率条 +verifiedAt）。render-spec.mjs是双模式：无 verdicts 出素 spec，有 verdicts 注入验收态。 - 回写边界：annotate 只写回
generator为specforge-draft*自产 spec，外部 generator（如 echo-spec）exit 3拒绝。schema/spec.schema.json的verdicts/verifiedAt是可选字段（素 spec 无此字段）。 - judge/fixer 分离红线：verify 回写改的是 spec 自身（需求的验收态），绝不改被验产物（代码/文档/配置）——运动员不当裁判。
校验
npm run check= schema 同步检查 + bundle 一致性检查 + version 一致性检查（三处 version）+node --test全部测试。
分发
- 两条通道并存、命名空间隔离：
npx skills add WangLiquan/specforge（skill 通道）与 Claude Code/plugin（plugin 通道，靠根目录.claude-plugin/{plugin.json,marketplace.json}）。加 plugin manifest 是纯 additive，不影响 skill 通道。发版流程：bumppackage.jsonversion →npm run build→ commit →git tag -a vX→git push --tags。
发布
showcase.html（spec 示例）/index.html（落地页）经 GitHub Pages 发布：branch deploy（main 分支 / 根路径），靠根目录.nojekyll原样托管静态文件，URLhttps://wangliquan.github.io/specforge/。无 Pages workflow。- README 首屏截图在
assets/spec-showcase.png，由 Playwright 截 showcase 页生成——改了 showcase 的视觉后需重新截图再同步 README，否则首屏与真实产物脱节。
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 · 31 lines · 918 tokens per session scan A 09c7234196e4
specforge CLAUDE.md is an instructions file published in the GitHub repository WangLiquan/specforge (2 stars, last pushed 2mo ago), licensed MIT. It adds 918 tokens to every session, about $0.0046 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
comet CLAUDE.md
Claude Code instructions for rpamis/comet, covering 回复语言, 测试, 提交前检查, commit 规范 and 项目结构规范.
superpowers-zh CLAUDE.md
Claude Code instructions for jnMetaCode/superpowers-zh, covering superpowers — 贡献者指南, 如果你是 ai agent, pull request 要求, 我们不会接受的内容 and 第三方依赖.
superpowers-zh GEMINI.md
Gemini CLI instructions for jnMetaCode/superpowers-zh, covering superpowers-zh 中文增强版, 核心规则, 可用 skills and 如何使用.
spec-superflow GEMINI.md
Gemini CLI instructions for MageByte-Zero/spec-superflow, covering spec-superflow and spec-superflow v1.2.0 | 阶段: {{state}} | 工作流: {{workflow}}.
web-dev-skills AGENTS.md
AGENTS.md instructions for timzaak/web-dev-skills, covering skill 编写规范, 规则归属, 编写 skill, 编排与角色边界 and 修改前后检查.
claude-skills CLAUDE.md
Instructions for mastepanoski/claude-skills, covering claude.md, repository purpose, testing skills locally, repository architecture and core structure.