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 skills add hxy91819/mason-skills --skill spec-leak-reviewgit clone --depth 1 https://github.com/hxy91819/mason-skillsWrote 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/skills/hxy91819/mason-skills/spec-leak-review)<a href="https://agentmods.dev/skills/hxy91819/mason-skills/spec-leak-review"><img src="https://agentmods.dev/badge/skills/hxy91819/mason-skills/spec-leak-review/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/skills/hxy91819/mason-skills/spec-leak-review"><img src="https://agentmods.dev/badge/skills/hxy91819/mason-skills/spec-leak-review.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 | $0.00060 | $0.02871 |
| Opus 5 | $0.00030 | $0.01435 |
| Sonnet 5 | $0.00012 | $0.00574 |
| Haiku 4.5 | $0.00006 | $0.00287 |
Grade A, and why
spec-leak-review 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 today.
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 — 108 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Spec Leak Review
只读审查,不修改被审文件。先按对象选择模式:
- 文本模式:目标是任意可读文本(包括
SKILL.md、AGENTS.md、CLAUDE.md、agents/openai.yaml、提示词、设计稿、配置、日志片段等),或用户明确要求审查文档内容。以文件或用户提供的原文为证据,不启动浏览器、不截图、不做视觉定级。 - 网站模式:目标是网站或网页应用,或用户明确要求检查网站页面的真实渲染。保留截图采集、bbox 定位和视觉复核;截图只服务于网站渲染证据。
混合输入按文件分别采集证据:文本文件走文本模式,网站页面走网站模式。报告开头写明模式和覆盖范围。
1. 确定受众与决策
先回答三件事并写进报告开头:谁会阅读或打开目标;带着什么疑问来;看完之后要做出什么决定或动作。文本模式中的“受众”可以是执行该文档的 agent、维护 Skill 的作者或文档读者;网站模式中的“受众”是页面用户。
受众决定一切判定——同一句“清单摘要 975d85…”对终端用户是噪音,对值班运维是工作对象。无法从用户输入、目标文件、网站上下文或代码判断受众时,先问用户,不要假设后展开审查。
完成标准:受众和“看完要做的决定”各能用一句话说清,且后续每条 finding 都能挂回其中一句。
2. 建立证据清单
文本模式
逐个读取用户指定的文本;没有指定文件时,只扫描任务范围内的文本,不把二进制或依赖目录当成证据。保留原文、文件路径和行号(必要时给出段落或字符范围),并记录文件用途、预期读者和是否由配置/模板生成。多文件交叉引用时以实际引用链为准,不凭目录名推断。
重点标出这些提示词残留:面向模型的会话开场、角色扮演、一次性用户请求、对当前对话的指代、隐藏推理或“请先……再……”的临时编排、把宿主/调用方式复述进正文的元说明,以及与文档任务无关的用户背景。提示词语气本身不是问题:Skill 的操作步骤、约束和安全门槛只要服务于预期读者,就属于正文。
完成标准:每个被审文件都有用途、读者、覆盖行数和证据位置;每条 finding 都能回指原文行号或明确标注为用户提供的片段。
网站模式
双轨采集,两轨都有时交叉核对。
截图轨:界面能跑起来时,用 scripts/capture-surface.js 采集每个网站页面,得到全页截图和带文档坐标 bbox 的可见文本清单。它把 placeholder、title、aria-label 一并收进来,并标出元素是否藏在 <details> 里、字号多大——这些是视觉复核的依据。脚本落地在浏览器错误页时以非零码退出,此时截图不能当作被审页面的证据。
拿不到浏览器或页面跑不起来时,用用户提供的网站截图逐区域盘点:标题、副标题、卡片标题、说明文字、字段名、字段值、状态标签、按钮、空态占位、时间戳。看到的每一块可见文字都进清单,包括看起来无害的;坐标从图上量出来。辨认不出的文字标注为不可辨认,不猜测内容。
代码轨:用截图里的文案原文在仓库检索定位组件与 i18n 条目,比按目录猜更可靠;每条给出 file:line。截图未出现的错误态、空态、tooltip、loading 等分支文案一并纳入清单。文案有配置或环境变量覆盖机制时,确认线上生效值,不拿源码默认值当结论。
截图里有而代码里检索不到的,标注来源为运行时数据或后端返回。部分区域因鉴权、外部网关或无法构造触发条件而采集不到时,如实记录范围和原因,不用源码推断冒充渲染结果。
完成标准:清单条目数不少于截图中可见文字块数;每条要么有 file:line、运行时数据或不可辨认说明,要么(网站渲染面条目)带 bbox。
3. 逐条判定
对清单每一条先归类,再给结论。以下类别适用于两种模式;“网站”措辞在文本模式替换为“该文档的预期读者”。
- 否定式免责——告诉读者“本页/本文不做什么”。信号词:不会、不再、不读取、不显示、不同步、仅、只。只有在回答隐私、计费、数据删除、权限等真实顾虑时才保留。
- 实现策略外泄——把取数、路由、优先级、回退、同步规则当成面向读者的说明;在 Skill 或工程文档中,若它是执行所需的规则则保留,否则视为重复或泄漏。
- 验收标准当文案——把“用于核对/用于确认 X 与 Y 是否一致”等测试目的写成面向读者的介绍。
- 元文案自述——“这里展示…”、“本页用于…”或解释文档/页面为何存在;职责已经由标题、结构或元数据表达时,通常删除或改写。
- 内部代号与技术标识——服务名、仓库名、组件名、hash、Release ID;受众不认识,或认识但无法据此采取动作。
- 无动作信息——只呈现状态不给下一步;异常时读者不知道能做什么;标识符不可复制。
- 占位与调试残留——
—、暂无、示例值、TODO、临时日志或调试输出随版本发布。 - 提示词残留(文本模式重点)——用户请求的口吻、会话上下文、模型角色/系统提示复述、一次性任务背景、隐藏推理或仅为引导 agent 当次作答的编排。判断依据是:删掉后是否损害文档作为长期规范/参考的可执行性;不损害则删除或改写。正常的规范性祈使句、步骤和安全门槛不因“像提示词”而标记。
What ships with it
4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- today Changed · +2 lines bc4eda16fdd8
- 8d ago First seen · 106 lines · 60 tokens per session scan A 76d4b0c5a25a
spec-leak-review is a skill published in the GitHub repository hxy91819/mason-skills (2 stars, last pushed today), licensed MIT. It adds 60 tokens to every session and 2,871 once invoked, about $0.0003 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 2026-09-04.
Other skills, from other repositories
webgl-holographic-foil
A self-contained WebGL2 hero: thin-film interference over a crushed-foil surface whose palette shifts with the viewing angle; move the cursor to tilt the film.
html-ppt-hermes-cyber-terminal
OpenDesign + BYOK: choosing and wiring your own model, hands-on — cost, quality, and the routing decision. Built as a decision-grade AI literacy deck for engineers, IT, applied-AI teams.
html-ppt-taste-brutalist
16:9 HTML deck in tactical-telemetry / CRT-terminal taste. Deactivated-CRT charcoal slides, white-phosphor monospace, hazard-red accent, scanline overlay, ASCII syntax, density over decoration. Distilled from Leonxlnx/taste-skill brutalist-skill (Tactical Telemetry mode).
visual-ralph
Visual Ralph orchestration for frontend UI from generated references, static references, or live URL targets, using $ultragoal with built-in visual verdict and pixel-diff evidence until the implementation matches and leaves a reproducible design system.
accessibility
Consolidated accessibility skill entrypoint for WCAG 2.2, ARIA Authoring Practices, cognitive accessibility, Section 508, EN 301 549, design intent verification, and the Accessibility Planner workflow.
make-resume
A Chinese-language tool for creating editable HTML resumes that can be changed in a browser and printed to PDF. It uses available resume templates when they are installed and otherwise provides a simpler fallback.