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/dora0512/interview-trainer/deep-reviewnpx skills add Dora0512/interview-trainer --skill deep-reviewgit clone --depth 1 https://github.com/Dora0512/interview-trainerWrote 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/dora0512/interview-trainer/deep-review)<a href="https://agentmods.dev/skills/dora0512/interview-trainer/deep-review"><img src="https://agentmods.dev/badge/skills/dora0512/interview-trainer/deep-review.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 | $0.00086 | $0.03319 |
| Opus 5 | $0.00043 | $0.01659 |
| Sonnet 5 | $0.00017 | $0.00664 |
| Haiku 4.5 | $0.00009 | $0.00332 |
Grade A, and why
deep-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 5d 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 — 161 lines — stays where its author put it; the contents beside it link to each section on GitHub.
渐进式深度复习 Skill / Progressive Deep Review
和 /review-skill 的区别：/review-skill 是「一次性快照卡片」用于快速回顾；/deep-review 是「跨会话渐进式学程」用于考前/弱点强化，每关用户先答 AI 后评，未通关自动倒灌成下次任务。
语言规则：指令用中文写，但用用户的语言主持学程和评分。
使用方式
/deep-review <技能编号或名称> # 启动 / 续上某技能的学程
/deep-review <技能> --only L4 # 只攻指定关
/deep-review <技能> --company <公司> # 用目标公司的追问风格做 L4/L5
/deep-review <技能> --reset # 重置该技能的掌握度档案，重新开始
核心原则（贯穿全流程）
- 不机械化：每步都结合「当前能力画像 + 最近真实面试 + 简历卖点」做个性化判断，不从 L1 死跑到 L5
- 用户先答：L1-L5 所有关都是用户先输出，AI 后评分。严禁在用户作答前给标准答案或暗示
- 跨会话续上：每个技能维护一份《掌握度档案》，启动时先读它，知道上次到哪关
- 低分倒灌：<3 分的关卡自动写入能力画像的薄弱点追踪表，下次启动诊断时优先排入
- 量化数据锁定：用户答 L3 项目落地时，AI 必须检查量化数据是否与
profile.md锁定数据一致
执行流程
Step 1：解析参数
兼容三种输入：技能编号（技能4/4）、技能名称、选项（--only L<N> / --company <公司> / --reset）。
从 knowledge-base/topics.md 模糊匹配技能；匹配不到时列出全部技能让用户选。
Step 2：现状诊断（90-120 秒）
强制读取（不读完不准进入下一步）：
| 文件 | 提取 |
|---|---|
knowledge-base/topics.md |
技能定义、能回答问题、达标判定、简历关联、项目实践、目标 L 级 |
data/capability-profile.md |
该话题当前水平、目标、上次测试、趋势、状态 |
data/capability-profile.md 「薄弱点追踪」表 |
过滤该技能相关薄弱点（注意是否已有 [deep-review-...] 标注） |
data/records/**/*.md（最近 3 场） |
「考察点总览」表中项目映射含本技能的题；Q 详情「发挥评估」表的 ❌/⚠️ 行 |
knowledge-base/star-stories.md |
通过技能定义的 STAR 引用找到关联 Story，提取锚点句和项目实战块 |
knowledge-base/deep-dive-questions.md |
抽取该话题的所有追问，按 L1-L4 分桶 |
knowledge-base/diagrams.md |
找该话题的 Mermaid 图 |
knowledge-base/analogy-bank.md |
该概念的生活类比 |
knowledge-base/methodologies.md |
用户自定义的方法论（评分时引用） |
knowledge-base/guides/*.md（若有对应文件） |
子系统/领域专题，提取关键章节作为评分锚点 |
data/deep-review-records/<技能号>-<技能名>-掌握度档案.md（若存在） |
上次每关得分、未通关记录、累计学时 |
data/pipeline.md |
--company 默认值：取当前 🟢 进行中且最近活跃的公司 |
Step 3：生成学程方案（用户可改写）
# 深度复习学程：技能 X — <技能名>
## 诊断快照
- 当前能力：能力画像 L<N>（上次测试 <日期>，趋势 <↑/→/↓>）
- 目标：L<目标>（按 topics.md 达标判定）
- 简历硬通货：<简历关联段的量化数据 1-2 个>
- 关联 STAR：<Story X>
- 真实面试历史：<公司>+<轮次> 在 <子点> 失分（来源路径）；未考过则注明"未被真实面试考察"
- 现有薄弱点：<从薄弱点追踪表过滤出的 1-3 条，标出现次数>
## 本次学程方案（预计 <X> 分钟）
| 关 | 状态 | 内容 | 出处 |
|----|------|------|------|
| L1 类比 | [跳过/✓ 攻] | … | 面试记录 |
| L2 原理图 | [✓ 攻] | 重点考 <子点> | 面试记录 + 追问清单 |
| L3 项目落地 | [✓ 攻] | STAR + 量化数据脱口而出 | STAR 库 |
| L4 抗追问 | [✓ 攻] | 模拟 <目标公司> 风格深挖 3-5 轮 | 追问清单 + 管线 |
| L5 跨场景迁移 | [✓ 攻] | 给新场景让用户迁移设计 | — |
| D1 <动态关名> | [✓ 攻] | <为何加 + 失分点出处> | 薄弱点追踪 / 面试记录 |
## 是否开始？
回复：`yes` 开始 / `改方案：<指令>`（如"跳过 L5"/"只攻 L4"/"换 <公司> 风格"/"缩到 15 分钟"）/ `预览 L4`
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.
- 5d ago First seen · 161 lines · 86 tokens per session scan A 0e5ccbd11e55
deep-review is a skill published in the GitHub repository Dora0512/interview-trainer (10 stars, last pushed 1mo ago), licensed MIT. It adds 86 tokens to every session and 3,319 once invoked, about $0.0004 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
interview-analysis
This skill should be used when the user asks to "/interview-analysis", "analyze my interview", "debrief my interview", "what did I miss in the interview", "review my interview transcript", or pastes a transcript / notes from a just-finished interview. Reads the transcript (pasted, from --from-file, or auto-pulled from…
case-practice
This skill should be used when the user asks for "/case-practice", "/pm-job-search:case-practice", "drill me on product cases", "practice PM case interviews", "MC case drill", or "test my case-interview recognition". Runs drill 1 of the case-practice methodology — a multiple-choice (MC) rapid-recognition drill that…
project-interview-skill
在项目根目录生成导学.md与面经.md；大厂工程向、第一人称口播（每题≥150字完整STAR）、 简历一句话简介、仓库相对路径阅读指引、量化合并文末（待测+如何测）。触发：面经、导学、 interview analyzer、项目分析、面试准备、STAR、简历亮点。.
rolecraft
JD-driven study companion. Paste a job description; get back what to study, what to build, and where else this role exists. Use when the user pastes a job description, says "process this", asks "what would this role take", or wants concepts, tech stacks, companies, projects, or a dashboard summary derived from past…
debrief
Post-interview review — reconstructs the Q&A from transcript or memory, captures interviewer intel and next-round forecasts, and turns the round into a revision list for the script. Use when the user says they just finished an interview, 刚面完 / 复盘一下 / 面试录音转文字给你 / 这轮被问了什么, or pastes an interview transcript or…
industry-brief
Generates industry and role reference reading (行业与岗位通识) for a target job — landscape, role expectations, must-know concepts, high-frequency interview topics, quotable viewpoints with sources — written to the workspace library/ folder. Use when the user asks for 行业通识 / 岗位认知 / 这个行业要懂什么 / 面这个方向需要补什么知识, or as step 4 of…