Backend Agent Resume Scout skill
A Codex skill that finds backend and agent projects on GitHub for a resume, checks them against source code and drafts CV text
- 375 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The skill clones third-party repositories to disk and runs its own search and PDF scripts
- Your stack and target position end up in search queries to GitHub and the web
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/backend-agent-resume-scoutDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skill clones third-party repositories to disk and runs its own search and PDF scripts
- Your stack and target position end up in search queries to GitHub and the web
This is third-party code. Review the repository files before installing.
What it does
Backend Agent Resume Scout picks GitHub projects that are worth putting on the resume of a backend developer or an AI-agent engineer and checks them against the source code. The skill is called `backend-agent-project-selector`. It first demands a mode: `agent-only`, `backend-only`, `mixed`, `safe-mode` or `challenge-mode`. Without a mode the search does not start. A script then looks for candidates on GitHub and the web, sorts them into directions (commerce, CRM and ERP, collaboration, tickets, knowledge bases, agent workflows) and drops demos, thin wrappers, frameworks, browser extensions, simple chatbots and shallow CRUD. Stars are a weak signal: a high rating adds no points by itself. You confirm a short list of three or four projects, then `pull_github_repos.py` clones the repositories, and conclusions rest only on local code and at least five pieces of evidence. Four files come out: `backend-agent-project-shortlist.md`, `backend-agent-project-resume-pack.md`, its PDF version and `repo-source-manifest.json`. The pack holds a short project description, five or six items on features and hard parts, interview questions and a rework plan.
Who it is for. Students and junior backend engineers who look for a resume project for an internship or a first job and want to answer questions about the source code with confidence.
Good fit when
- You need a resume project for a backend internship that goes beyond a typical online store with CRUD
- An agent project on the resume should include data, state transitions and value for users
- You have two weeks to prepare and the `safe-mode` mode with simple deployment fits
- You want answers ready for interviewer questions about states, errors and code branches
- You have chosen Java, Python, Go or Node.js and look for a project in your stack
Not a fit when
- You are interested in IoT, embedded systems or hardware: such projects are downweighted by default and need an explicit request
- The project is already chosen and you only need resume text: most of the skill's work goes into search and cloning
- You need an answer in a minute: the skill stops at the mode choice and at the short list and waits for your confirmation
Example request
Use $backend-agent-project-selector, mixed mode: find one agent project and one Java backend project for an internship resumeLimitations
The skill is written for Codex: installation goes through `~/.codex/skills` and the `agents/openai.yaml` file targets that agent. The author did not test other agents.
The instructions and references in `references/` are in Chinese, so ask for an English or Russian answer separately.
The skill clones the repositories it finds and runs its own search and PDF scripts, so you need network access, `git` and Python. The resume is built from what the skill found in the code. Reread it before sending it to an employer: the wording must match your real contribution.
How to disable. Delete the `~/.codex/skills/backend-agent-project-selector` folder. The cloned repositories and result files stay where the skill wrote them and are removed separately.
FAQ
How do I install the skill in Codex?
Clone the repository, run `mkdir -p ~/.codex/skills` and `cp -R backend-agent-project-selector ~/.codex/skills/`. In Codex, call the skill with `Use $backend-agent-project-selector ...`.
Which search modes exist?
Five: `agent-only` returns agent projects only, `backend-only` classic backend, `mixed` one project of each kind, `safe-mode` projects that are easy to deploy, and `challenge-mode` harder and more distinctive ones. A mode is mandatory.
What does the skill do if no mode is given?
It lists all modes and asks which one to use. Search and recommendations start only after you answer.
Which files appear after the run?
The working folder holds `backend-agent-project-shortlist.md` with the candidates, `backend-agent-project-resume-pack.md` with the final pack, its PDF and `repo-source-manifest.json` with the cloning results.
Why does the skill not rely on star counts?
A rating pushes the choice toward popular, similar projects. So the skill searches the middle range and niche directions too, and counts stars as a weak signal.
Discussion
No questions or reviews yet.
Related
SkillsJustHireMe is a local-first job-search desktop app that gives an agent a skill and a stdio MCP server to work with job data safely.
A skill that builds a Chinese interview prep pack from a resume and a loose role direction, using recent candidate reports.
cafe3310's personal skill collection: spec-driven project work, reviews, reports, browser, 3D and video in one repository
Project Interview Skill turns a repository into a source-code course with resume points and interview questions.
get-job.skill (实习.skill)
get-job.skill
A skill that prepares a job search: it researches the role, rewrites your resume and drills you on every bullet
A skill set for resumes and job search: tailoring to a vacancy, ATS pass-through checks, cover letters and interview preparation
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Backend Agent Resume Scout (this card) | no | 375 | |
| Resume Skills | no | 3,153 | rewrites a resume for a vacancy and checks it for ATS, while this skill picks the project itself and verifies it against source code |
| Interview Skills | no | 402 | runs an interview on a vacancy and resume, while here questions are prepared for the chosen project |
README in short
The project text in Chinese and English opens with the problem: a resume with projects from rankings does not survive questions about the code. It then covers features, modes, a flow of confirmable steps, call examples, the output files and the repository structure.
README badge
Original SKILL.md
--- name: backend-agent-project-selector description: "Find suitable traditional software backend or business-grade AI agent projects from GitHub for students or early-career engineers, evaluate resume value, and generate a Markdown resume-writing package. Use when a user asks to choose backend projects, Java/Spring projects, distributed-system projects, complete AI agent business projects, multi-agent systems, LLM apps, GitHub projects for resumes, or turn selected projects into resume bullets/interview talking points. Supports modes: agent-only, backend-only, mixed, safe-mode, and challenge-mode. If no mode is specified, ask the user to choose before searching or recommending. Backend recommendations exclude IoT, embedded, hardware-integration, device-management, and industrial-control platforms unless requested. Avoid simple browser extensions, thin LLM wrappers, shallow AI plugins, and browser-automation libraries unless requested." --- # 后端 / Agent 项目选择器 ## 能力范围 从 GitHub / Web 搜索适合写进简历的后端项目或完整业务型 Agent 项目，筛选后生成 `backend-agent-project-resume-pack.md`。 推荐模式是必填项。用户必须指定 `agent-only`、`backend-only`、`mixed`、`safe-mode`、`challenge-mode` 中的一种；如果用户初始化找项目、选择项目、推荐项目或生成简历项目但没有指定模式，先完整列出所有可选模式并让用户选择，不要开始搜索或生成推荐。 ## 必须遵守 - 用户每次初始化找项目、选择项目、推荐项目或生成简历项目时，如果推荐模式缺失，必须先完整列出所有可选模式并反问，不能自行默认、推断或继续执行。 - 除非用户明确要求不要联网，否则要搜索当前 GitHub / Web 项目后再推荐。 - 优先完整业务系统，避免 demo、插件、薄封装、纯框架和只会调用一次大模型的项目。 - `backend-only` 不要默认只推荐 API 网关、IAM、调度、监控等底层技术项目；除非用户明确要求基础设施 / 中间件，否则应优先覆盖电商交易、内容社区、协作办公、工单客服、CRM/ERP、知识库、网盘、在线教育等完整业务项目。 - 不要避开有业务闭环的业务项目；只避开浅层 CRUD、教程复刻、没有状态流转 / 异步链路 / 失败恢复 / 权限边界的业务项目。 - 默认不要把 IoT、嵌入式、硬件接入、设备管理、工业控制类平台作为后端项目推荐；除非用户明确要求硬件 / IoT / 物联网方向，否则这些项目应降权或淘汰。 - 先构建多样化候选池，再做最终选择；不要直接推荐全局高星项目，也不要把 star 数作为主要排序依据。 - star 数只作为“项目已有一定社区验证”的弱信号：达到 1k star 即可纳入正式候选池并认真评估；超过 1k 后不应因为 star 更高而显著加分。 - 搜索候选时必须主动覆盖中等 star 区间和细分领域项目，避免只用 `sort=stars` 或只看几万 star 项目；后端和 Agent 项目都适用该规则。 - 候选池阶段要读取 README 前若干行做项目类型 probe，用于判断是否为业务系统、框架 / SDK / 桌面壳 / 工具链；README probe 只能用于筛选和项目定位，不能支撑最终“负责功能 / 技术难点”。 - 构建候选池时优先使用 `references/search_github_candidates.py` 自动搜索、README probe、去重、分桶和初筛；如果手动搜索，最终也要输出同等字段的候选池和短名单确认内容。 - 拉取最终候选源码前，必须先向用户展示 3-4 个短名单项目、每个项目的选择理由和主要淘汰理由，并等待用户确认方向；用户确认后才能执行 `pull_github_repos.py`。 - 最终入选项目前必须运行 `references/pull_github_repos.py` 真实拉取对应 GitHub 仓库到本地；未经过该脚本成功拉取并写入 manifest 的项目不得进入最终推荐。 - 源码验证只能基于 `pull_github_repos.py` 拉取到本地的仓库目录进行；不能用 GitHub raw/API、README、网页搜索、模型记忆或经验判断替代本地源码验证。 - 如果脚本执行失败、仓库无法拉取、manifest 中该仓库状态不是 `cloned`、本地源码不可读，或无法从源码提取至少 5 个证据点，该项目必须淘汰。 - 必须区分 `已有能力`、`建议改造`、`可写入简历`。 - 简历功能点默认输出为“建议简历功能点（完成对应改造后可写）”；除非用户明确已经实现或本轮完成代码改造，否则不能把建议改造写成已完成成果。 - 项目亮点必须挖掘技术难度，不能只写“实现功能 / 接入组件 / 提供接口”；优先挖数据同步与一致性、MQ 异步链路、缓存与高并发、并发控制与幂等、任务调度、线程池与异步编排、流量治理、数据库优化、检索索引、权限安全、可观测性、接口治理、规则引擎、状态机、文件处理、实时通信、交易链路、Agent 工程等机制。 - 默认使用中文输出，除非用户要求其他语言。 ## 参考资料加载 按任务需要加载，不要一次性加载所有资料： - `references/执行流程.md`：项目选择任务必载。 - `references/用户输入模板.md`：用户需要模板、示例，或推荐模式缺失需要反问时加载。 - `references/筛选评分.md`：筛选、分桶、推荐模式打分时加载。 - `references/简历写法.md`：写简历条目和面试问题前加载。 - `references/输出模板.md`：生成最终 Markdown 文件前加载。 - `references/search_github_candidates.py`：候选池搜索和短名单确认前优先执行；用于自动搜索 GitHub、读取 README probe、去重、分桶、排除前端 / 库 / IoT / 框架 / SDK / 桌面壳 / coding-agent 工具链等不合适项目，并生成候选池 JSON 与短名单预览 Markdown。