adversarial-review SKILL.md skillA
adversarial-review from sichenai/sichen-skills is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.
mdr add sichenai/sichen-skills/adversarial-review@v1.1.1Pins v1.1.1 in mdr.lock. Pin the fingerprint instead.
- Audit grade
- A, 17 of 17 checks passed
- Kind
- SKILL.md skill
- Repository
- sichenai/sichen-skills
- Path
skills/adversarial-review/SKILL.md- Versions recorded
- 3
- Latest version
- v1.1.1, committed Oct 12, 2026
- First version recorded
- Aug 17, 2026
- Size
- 14,009 bytes
- Repository stars
- 3
- License
- MIT
以攻击者视角审查 AI 交付的方案/设计/代码（AI 审 AI）。当用户说"对抗性审查""验收一下""这个方案靠谱吗""AI 写的这个能信吗""帮我挑刺""adversarial review""红队审查"，或在编程/写作 agent 完成任务需要验收时触发。先核验交付真实性（D0：需求符合性矩阵、声明证据链、幻觉依赖），再做安全/正确性/健壮性/现实性四维破坏测试，输出带攻击路径、风险等级（P0-P3）与置信度的报告。注意：「这个怎么样」类随口询问不触发本 skill。.
Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.
What the file says
# 对抗性审查（Adversarial Review） ## 0. 核心定位 当一个 AI（编程 agent、写作 agent、方案 agent）交付了代码/方案/设计，在用户验收之前，由对抗视角做系统性破坏测试——**既攻击交付物本身，也攻击交付过程中的声明**。 审查对象是**三方整体**，不是孤立的交付物： ``` ① 原始需求 —— 用户当时到底要什么（原始 prompt / 需求描述 / 任务书） ② 交付物 —— AI 实际产出的东西（代码 / 方案 / 设计 / 文档） ③ 完成声明 —— AI 声称自己做了什么（"已完成""已测试""修复了"） ``` - ② vs ③ 对照 → 查虚假声明（F3） - ① vs ② 对照 → 查需求缩水（F2） - ① vs ③ 对照 → 查承诺落空 三方缺失时不臆测、不阻塞：相应检查项降级为「未覆盖」，在报告盲区段声明。 **四条边界（违反任何一条即跑偏）：** 1. **不替代验收决策**——输出风险清单，用户做最终判断 2. **不做建设性补全**——只攻击和暴露，附修复方向但不重写 3. **不审查「能力边界」本身，只审查「证据链」**——能力是运行时配置（模型×工具×connector×权限），动态变化；声称做过 X 就要拿出做 X 的中间产物，拿不出一律按「未验证声明」处理（详见 D0） 4. **不追求零问题**——价值在于把问题按风险显性化；找不到高等级风险就明说，**禁止凑数** ## 1. 触发判定 | 触发 | 不触发 | |---|---| | "对抗性审查""验收一下 agent 写的代码""这个方案靠谱吗""AI 写的这个能信吗""帮我挑刺""红队审查" | 「这个怎么样」「看看如何」类随口一问（重型流程不被无意激活） | | agent 完成任务后用户要求验收 | 用户自己写的东西要求 review 情绪/风格 | | 贴入其他 AI 工具的产出要求把关 | 用户明确只要夸奖或只要执行 | ## 2. 审查流程（七步，顺序不可乱） ### 第一步：三方输入收集 - 当前会话内的 agent 任务：直接从对话上下文提取三方输入 - 跨工具场景：要求用户贴入；用户只给交付物时如实标注"只有交付物" - 原始需求缺失时先问用户能否提供；不能提供则记录为盲区，继续 ### 第二步：建立攻击地图 列出四张清单： 1. 需求原子条目——**逐字引用用户原文，禁止改写式转述**（改写会把审查方自己的解读混入需求，同会话场景下这是 F2 漏检的主通道） 2. AI 的声明清单（所有"已完成/已测试/已修复"） 3. 外部依赖清单（API、库、文档、数据源） 4. 交付物结构与信任边界 …
Read the whole file at its exact version.
How to install
mdr add sichenai/sichen-skills/adversarial-review@sha256:3db1c0453213b033The line at the top of this page pins the label v1.1.1, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.
[](https://markdownregistry.com/a/art_vdloblrjgc2iz2mb)
2 badge image fetches in 30 days. What the badge shows.
Versions
The registry has recorded 3 versions of this file, between Aug 17, 2026 and Oct 12, 2026, about one every 28 days.
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| v1.1.1 latest | 2026-10-12 | bb5df4d | 14,009 B | A | view · diff |
| v1.1.1 | 2026-08-25 | 0dee6bf | 14,013 B | A | view · diff |
| v1.1.0 | 2026-08-17 | 59946c2 | 13,354 B | A | view |
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (14009 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
sichenai/sichen-skills · 3 stars · license MIT · pushed 2026-10-27 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_vdloblrjgc2iz2mb GET https://markdownregistry.com/api/v1/resolve?ref=sichenai/sichen-skills/adversarial-review GET https://markdownregistry.com/api/v1/blob/3db1c0453213b033a54c25745efd986a31049c3596c7609ce131a96e0cb8a9cc
Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.
More from sichenai/sichen-skills
Every file in sichenai/sichen-skills