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/ranxi2001/zero2agent/classify-interview-questionsnpx skills add ranxi2001/zero2Agent --skill classify-interview-questionsgit clone --depth 1 https://github.com/ranxi2001/zero2AgentWhat 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 | $0.00104 | $0.05226 |
| Opus 5 | $0.00052 | $0.02613 |
| Sonnet 5 | $0.00021 | $0.01045 |
| Haiku 4.5 | $0.00010 | $0.00523 |
Grade B, and why
classify-interview-questions scanned grade B with 1 finding 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
`--llm` 默认从 Codex 配置读取连接信息：优先显式 `--codex-config` 或 `~/.codex/config.json` / `codex.json`，否则读取官方 `~/.codex/config.toml` 当前 `model_provider` 和 `auth.json`。兼容 `baseURL/base_url`、`token/apiKey`，按 `wire_api` 调用 Responses 或 Ch How it starts
The opening of the file, as written. The whole thing — 296 lines — stays where its author put it; the contents beside it link to each section on GitHub.
classify-interview-questions：面试题分类分发
将新面试题按考察维度分类，拆分后加入已有题库。不新建独立面经实录文章。
两个目标仓库
- Agent、LLM、RAG、训练、多模态、AI Coding 与 AI 工程题：当前
zero2Agent/learn-agent-interview/。 - Java/Go/Python、JVM、并发、操作系统、网络、数据库、缓存、消息队列、分布式、前端与通用工程八股：相邻
../zero2Leetcode/_includes/interview-seasons/2026/summer.md。 - 算法/手撕题只进入 zero2leetcode 的算法题单，不计入传统八股总数；题干不完整时排除，不补造。
- 机器召回使用本 Skill 的
question-index.json；question-index.md是人工维护源。每次修改 Markdown 后必须重建 JSON 并运行 stale check。 question-frequency.json是 Agent 题单频次事实源。frequency等于evidence中可归因的独立面经出现次数；题库汇总、“高频题”等不可追溯标签不计数。正文只在现有最小主题组内按频次降序排列，同频按稳定的firstSeenOrder排列。- 两个仓库分别检查 worktree、计数、提交和推送。不得把一个仓库的 commit 混入另一个。
来源与超链接契约
来源链接是题库可信度的一部分。以后新增题、增强已有题和调研答案都按以下标准维护，不只写不可核验的公司名或“高频题”。
面试题证据
-
有公开原文时，正文
> 来源：使用可点击的 Markdown 链接，链接文字保留公司、岗位和轮次，例如：> 来源：[阿里云 Agent Infra 一面](https://www.nowcoder.com/feed/main/detail/<id>) -
链接必须直达原始面经文章或用户指定的一手页面，不使用搜索结果页、信息流首页、聚合转载、短链或无关主页。多个独立来源分别给链接，不能把多个 URL 藏在一个笼统标签里。
-
同一来源要贯穿三处：正文来源行、
question-index.md的题目来源、question-frequency.json的 evidence。人工索引保留 Markdown 链接；频次 JSON 中每个独立原文 URL 单独占一个 evidence，同一核心题里的同一 canonical URL 只计一次。 -
增强已有题时保留原有链接，再追加新来源或追问链接；不要把已经可点击的来源降级成纯文本。
-
URL 只是可追溯入口，不自动证明内容是一手面经。仍需按真实面试过程、公司/岗位/轮次和问题上下文判断是否可归因，转载、题库汇总和营销内容仍不计频次。
-
用户明确授权的本地题库或私有材料可以作为来源，但没有公开 URL 时必须写明“用户授权题库（无公开 URL）”，不得伪造链接，也不得把本机绝对路径、私有下载地址或凭证写进站点。若能定位到公开原文，先验证内容一致再补回原始链接。
-
无公开 URL、又没有用户明确授权或其他可归因证据的题目，不进入频次事实源。
答案事实依据
- 协议、框架、模型、Kubernetes、AIOps 等会变化或容易混淆的事实，优先查当前一手资料，并在支持该结论的正文附近直接链接官方文档、官方仓库、标准或论文，不只在文末列一个泛化“参考资料”。
- 链接应指向支持具体结论的页面；官方资料没有直接支持时，明确标注这是工程推断或经验判断，不借链接制造过度确定性。
- 答案实质依赖外部资料时同步更新
THIRD_PARTY_NOTICES.md。独立改写答案，不复制外部文章的长段落、图表或受版权保护内容。
链接验收
- 发布前验证新增链接可访问、页面标题与标注一致、没有误链到评论区或推荐页；登录受限页面至少确认 URL 是原始文章地址。
- 核对每个新增或增强题的正文来源、人工索引和频次 evidence，不允许其中一处有 URL、另两处丢失 URL。
- 汇报时增加来源覆盖：公开原文链接数、用户授权但无公开 URL 的来源数、无法归因而排除的来源数。
维度分类表
| 编号 | 维度 | 目录 | 典型考察内容 |
|---|---|---|---|
| 01 | 架构选型 | 01-architecture-design/ |
ReAct/Plan-Execute/ToT、Agent 组成、设计范式、规划器 |
| 02 | 工具管理 | 02-tool-management/ |
参数校验、工具路由、多工具调度、Mock 生成 |
| 03 | 容错与鲁棒性 | 03-fault-tolerance/ |
超时处理、误操作防范、幻觉治理、失败恢复 |
| 04 | 记忆与上下文 | 04-memory-context/ |
长对话、模糊需求、上下文污染、长短期记忆、to-do list |
| 05 | 评估与全局观 | 05-eval-and-vision/ |
量化评估、落地挑战、AI 工具价值/边界、行业认知 |
| 06 | 多智能体协作 | 06-multi-agent-collab/ |
角色分工、通信机制、冲突仲裁、记忆共享 |
| 07 | 工程化踩坑 | 07-engineering-pitfalls/ |
死循环、状态丢失、成本控制、AI Coding 实践、工具使用 |
| 08 | Prompt 工程 | 08-prompt-engineering/ |
模板构建、Skills 机制、好/差 Prompt 区别、框架创新 |
| 09 | RAG 与检索 | 09-rag-retrieval/ |
chunk 设计、查询改写、召回精排、Embedding/ReRank 微调 |
| 10 | 训练与模型 | 10-training-and-data/ |
数据清洗、LoRA、PPO/DPO/GRPO、位置编码、归一化、量化部署、多模态 |
| 11 | AI 代码测试 | 11-ai-code-testing/ |
覆盖率插桩、前置分析、代码过滤 |
| 12 | 业务 AI 工程 | 12-business-ai-engineering/ |
业务需求拆解、方案选型、效果评估、智能客服与业务落地 |
| 13 | 简历项目拷打 | 13-project-deep-dive/ |
项目部署、框架选型、意图识别、工具设计、知识库构建、性能优化 |
| 14 | 公司偏好（派生页） | 14-company-preferences/ |
从各维度来源统计公司考察偏好，不直接写入新题 |
| 15 | Agent 概念 | 15-agent-concepts/ |
Harness/Context Engineering、Vibe Coding、MCP、Skills 等概念辨析 |
| 16 | Agent Infra | 16-agent-infra/ |
Runtime、Checkpoint、幂等、Sandbox、Kubernetes、调度与可观测 |
| 17 | AI Infra | 17-ai-infra/ |
分布式训练、LLM Serving、GPU 调度、模型发布与 AIOps |
What ships with it
13 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.
- question-frequency.json 185 KB
- question-index.json 277 KB
- question-index.md 92 KB
- scripts/build_question_index_json.py 4.1 KB runs code
- scripts/extract_and_recall.py 31 KB runs code
- scripts/question_frequency.py 6.5 KB runs code
- scripts/recall_similar_questions.py 13 KB runs code
- scripts/sort_questions_by_frequency.py 11 KB runs code
- scripts/sync_question_frequency.py 4.0 KB runs code
- tests/test_extract_and_recall.py 7.9 KB runs code
- tests/test_question_frequency.py 7.5 KB runs code
- tests/test_recall_similar_questions.py 5.5 KB runs code
- tests/test_sort_questions_by_frequency.py 2.3 KB runs code
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 · 296 lines · 104 tokens per session scan B f156eed3b457
classify-interview-questions is a skill published in the GitHub repository ranxi2001/zero2Agent (367 stars, last pushed 3d ago), licensed MIT. It adds 104 tokens to every session and 5,226 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
kungfu-agent-onboarding
Use when a user asks to understand, start, inspect, extend, or safely operate installed Kungfu; verify the installed pack, select one intent route, personalize the explanation, and propose one smallest safe next action.
llmtornado-tutorial-generator
Generates comprehensive code tutorials on LlmTornado API formatted for Medium publication with examples, explanations, and best practices.
manim-video
Manim CE animations: 3Blue1Brown math/algo videos.
hive.worker-delegation
Concrete patterns for breaking colony work into parallel worker jobs via runplaybook — when fan-out helps, how to model the goal as a tracker table, write the worker skill, author the playbook, pilot, and let convergence retry/resume the gap.
hive.x-automation
Read before automating X / Twitter with browser tools. Verified flows for post, reply, delete, search-and-engage, plus the Draft.js compose quirks that silently disable the send button. Includes the daily-reply and job-market-reply playbooks. Requires hive.browser-automation for the underlying screenshot + coordinate…
hive.chart-creation-foundations
Required reading whenever any chart tool is available. Teaches the one-tool embedding contract (call chartrender → live chart appears in chat AND a downloadable PNG lands in the queen session dir), the ECharts (data viz) vs Mermaid (structural diagrams) decision, the BI/financial-grade aesthetic baseline (no…