ResumeSkills: Markdown Skill Files That Teach Claude Code Resume Workflows

What ResumeSkills Actually Is: Prompt Files, Not a Program

The README states the premise directly: skills are markdown files that give AI agents specialized knowledge and workflows for specific tasks. There is no runtime, no parser, no scoring engine. When you install this repository you are copying text into a folder that Claude Code, Cursor, Windsurf, Codex or Gemini CLI scans for skill definitions, and the agent then decides on its own when a resume task matches a skill. The repository's primary language is listed as unknown, which fits a project whose payload is prose. That distinction matters for how you evaluate it. You cannot unit test a markdown file. You can only read it and judge whether the workflow it describes is one you want an agent to follow on your behalf. The audience is narrow and specific: people who already work inside an agentic coding tool and want their job search handled through the same interface. Anyone expecting a web app with a drag-and-drop resume editor is looking at the wrong repository.

The 20 Skills and the Categories They Cover

The README groups 20 skills into five categories. Resume optimization covers resume-ats-optimizer, resume-bullet-writer, resume-quantifier, resume-formatter and resume-section-builder. Job search strategy covers job-description-analyzer, resume-tailor, resume-version-manager and offer-comparison-analyzer. Supporting documents covers cover-letter-generator, linkedin-profile-optimizer, portfolio-case-study-writer and reference-list-builder. Interview and negotiation covers interview-prep-generator and salary-negotiation-prep. Specialized roles covers tech-resume-optimizer, executive-resume-writer, academic-cv-builder, creative-portfolio-resume and career-changer-translator. The split is sensible. A career changer and a tenured professor need different framing, and the repository does not pretend one skill handles both. Two entries deserve scrutiny. resume-quantifier is described as finding opportunities to add metrics and estimating when numbers are unknown, which is a workflow that can produce invented figures if the agent is not constrained, so read that file before trusting it. offer-comparison-analyzer claims total compensation analysis, which is a structured comparison task an agent can do reasonably well from pasted offer letters.

How the Agent Picks a Skill From a Plain Request

The mechanism is routing by intent, and the README shows it with examples. Typing optimize my resume for ATS maps to resume-ats-optimizer. Typing improve my resume bullets maps to resume-bullet-writer. Pasting a job description after should I apply to this job maps to job-description-analyzer. Prep me for an interview at Google maps to interview-prep-generator. The agent reads the skill files available in its skills folder, matches the request against the descriptions, and loads the relevant instructions into context. This is why installation location matters. The CLI install with the -g flag places skills globally so they work across all projects, while the same command without -g installs into the current project only. The manual path in the README copies everything into ~/.cursor/skills, and the uninstall examples reference both ~/.agents/skills and ~/.cursor/skills, which tells you the two directories the project expects to be in play. If you install globally and later wonder why an unrelated project is behaving differently, the global skills folder is the first place to look.

Installing and Removing Skills With the npx CLI

The README gives three installation routes. The recommended one is the CLI: npx skills add Paramchoudhary/ResumeSkills -g -y installs all 20 skills globally, and dropping the -g flag installs to the current project. Verification is npx skills list for project scope and npx skills list --global for the global set. The manual route clones the repository and copies the skills directory into an agent's skills folder with mkdir -p ~/.cursor/skills followed by cp -r ResumeSkills/skills/* ~/.cursor/skills/. The third route is downloading individual skill files from the /skills directory and placing them by hand. Removal is per-skill, for example npx skills remove resume-ats-optimizer, or bulk through rm -rf ~/.agents/skills/resume-* and the equivalent under ~/.cursor/skills. Note that the bulk removal depends on every skill directory starting with the string resume-, which is not true for job-description-analyzer, cover-letter-generator, linkedin-profile-optimizer, interview-prep-generator, salary-negotiation-prep, offer-comparison-analyzer, academic-cv-builder, reference-list-builder, portfolio-case-study-writer or career-changer-translator. Those ten will survive the wildcard and need explicit removal.

The ATS Claims and Why the Numbers Do Not Hold Up

The README's Why These Skills Matter section asserts that 75% of resumes are rejected by ATS before humans see them, that the average job gets 250 applications, that users see 2-3x more interviews per application, that the job search shortens by two months on average, and that negotiations yield $10K+ higher offers. None of these figures carry a source in the supplied material, and the repository contains no benchmark harness, no test corpus and no evaluation script that could produce them. Treat them as marketing copy rather than measurement. The same caution applies to the ATS optimization skill itself. Applicant tracking systems vary widely in how they parse and rank documents, and a markdown instruction file cannot query a specific vendor's parser. What resume-ats-optimizer can plausibly do is apply general heuristics about keyword matching and formatting conventions, which is useful but is not the same as a compatibility score. If your decision depends on a hard number, this project does not supply one you can defend.

Where the Skill-File Model Breaks Down

The first limitation is that nothing here is versioned in a way that protects you. The repository shows no retrieved releases, so there is no changelog to read before an update changes how a skill behaves. A global install with -g means one pull can alter the instructions your agent follows in every project on the machine, with no diff review unless you check the skills folder yourself. The second limitation is the honesty boundary. resume-tailor is described as customizing a resume for a specific posting while maintaining truthfulness, and resume-quantifier estimates numbers when real figures are unknown. Those two goals pull against each other, and the enforcement lives entirely in the wording of the markdown. An agent will do what the file says, not what you meant. The third limitation is scope. This is not a tool for someone who has never written a resume and wants output they can send unread. It assumes you have a resume, a target role and the judgement to reject a bad suggestion. Used as a first draft generator with no review, it will produce confident text that may not be true.

How This Differs From a Hosted Resume Platform

The obvious comparison is a hosted resume service such as Rezi or Teal, which combine a document editor, a job tracker and an ATS scoring feature behind a web account. The difference in approach is structural. A hosted platform owns the document, applies its own scoring model and stores your data on its servers. ResumeSkills owns nothing. Your resume stays in whatever file you paste it into, the analysis happens inside your agent session, and the output is text you copy back out. That means no account, no subscription and no vendor holding your employment history, which is a real advantage for anyone who dislikes uploading a career to a third party. It also means no job board integration, no application tracking dashboard and no persistent record unless you build one. resume-version-manager is described as tracking different resume versions and maintaining a master resume, but the storage is whatever your agent can see in the working directory. If you want the tracker, you maintain the files.

Licence, Maintenance and What to Check Before You Commit

The repository is MIT licensed, which permits commercial use, modification and redistribution provided the copyright notice and permission notice are preserved. That is permissive enough to fork the skills into an internal hiring or career-coaching workflow, and permissive enough to ship them inside a paid product. It also means no warranty of any kind, which for a project made of instructional text is the honest position: the authors are not guaranteeing that resume-ats-optimizer improves your callback rate. The last push date is 2026-06-19, so the project was active recently, but with no releases retrieved there is no stable tag to pin to. If you install globally, record the commit you installed from, because npx skills add pulls the current state of main and a later update will silently change the instructions. For a team, the safer pattern is a project-scoped install without -g so the skills live alongside the repository that uses them and are reviewed like any other file.

Editorial conclusion

Adopt ResumeSkills if you already drive Claude Code, Cursor or a similar agent from a terminal and you want structured prompt scaffolding for resume, cover letter and interview work, and if you accept that the skills are instruction files rather than tested software. Do not adopt it if you need a validated ATS score, a hosted resume builder, or anything with a support commitment, because there is no code to test and no release history. Before installing, open two or three files under /skills, for example resume-ats-optimizer and resume-quantifier, and check whether the instructions match your own standards on truthfulness and metrics, since the agent will follow whatever those files say.

Official sources

1. Issues
2. License: MIT
3. Paramchoudhary/ResumeSkills on GitHub
4. README