career-ops: a Claude Cowork plugin that scores job postings and tailors resumes
Claude Cowork plugin for job seekers. 9 AI skills: evaluate job postings, generate ATS-optimized resumes, scan company career portals, track applications, draft outreach. Works in any industry.
At a glance
- What is it?
- career-ops packages nine Claude Cowork skills for job seekers, from A-F posting evaluation to ATS resume tailoring. The README documents the skills and the install path, but leaves the skill internals and the data directory's real boundaries unexplained.
- Who is it for?
- career-ops fits people running a structured search across many postings who already use Claude Cowork and are comfortable pasting a resume into a local data directory. It is the wrong tool if you want a hosted dashboard, if you need the plugin to submit applications unattended, or if you expect the repository to be stable: the repository name itself warns against forking while v2 is being updated.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 55 days ago.
- What is it written in?
- Mainly HTML, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
What career-ops solves for job seekers
Job searching at volume produces a specific kind of mess. You have a folder of postings, a resume that gets rewritten for each one, a spreadsheet of applications, and no consistent way to compare opportunities. career-ops is an attempt to put that workflow inside Claude Cowork as a plugin with nine named skills. The README frames it as turning your job search into a system, and the skill list is where that claim has to hold up: evaluate, tailor-resume, scan, triage, track, apply, research, outreach, and compare.
The audience is broad by design. The README states it works for any industry, and the repository backs that with 15 archetypes, including Technology, Finance, Healthcare, Legal, Creative/Marketing, Operations, Sales/BD, Education, Executive, Trades, Customer Success, People/HR, Government/Nonprofit, Scientific/R&D, and Non-Software Engineering. Each archetype adjusts scoring weights and evaluation language. That is more than a marketing line: a trades posting and a research scientist posting do not share evaluation criteria, and the archetype list is the mechanism for handling that.
It is adapted from santifer/career-ops, a detail worth knowing because the upstream project is the origin of the skill design. If you have seen the original, this repository is the Claude Cowork port, not an independent build.
How the nine skills and the A-F evaluation actually work
The plugin is a directory of skills, agents, commands, and references, with a .claude-plugin/ manifest at the top level. Claude Cowork loads it, and the skills become things you invoke by saying them in natural language. There is no CLI to learn and no config file to write for the core loop.
The evaluate skill is the one with documented structure. You paste a job posting as text or a URL, and the assessment comes back in six labeled parts: A. Executive Summary (archetype, seniority, one-line verdict), B. Background Match (every JD requirement mapped to your experience), C. Positioning Strategy, D. Compensation & Market, E. Tailoring Plan, and F. Interview Prep (STAR stories mapped to JD requirements). The score runs from 1.0 to 5.0, and the README explicitly says it is honest, not inflated. Section B is the part that matters most in practice: mapping every requirement to your background is what surfaces the gaps you would otherwise discover in the interview.
The data flow is local. Your profile, applications, and resumes live in a data/ directory that .gitignore excludes. The README notes that nothing is sent to external services beyond what Claude uses to help you, and names two exceptions: web searches for salary data and ATS API calls for job scanning. So the scan skill is the one that reaches outside your machine, and the README says so rather than claiming total isolation.
A caveat the README does not address: it documents what each skill does in a table, not how the skills are implemented. The skills/ directory holds that, and the README does not reproduce it. If you need to know whether evaluate reads your whole profile or a summary, the README is silent, and you have to read the skill files.
Installing career-ops and running a first evaluation
The README gives two install paths. The first runs the plugin from a local directory during development, which is the fastest way to try it without touching your plugins folder:
claude --plugin-dir ./career-ops-pluginThe second clones the repository into your plugins directory. Note that the README's clone URL and the repository name differ; the README uses andrewshwetzer/career-ops-plugin, while the repository you are reading is andrew-shwetzer/career-ops-plugin-do-not-fork-currently-updating-v2-. Use the URL from the README if you want the path it documents, and confirm the remote resolves before you rely on it:
git clone https://github.com/andrewshwetzer/career-ops-plugin.gitOnce installed, the README's Quick Start is five steps. First, say "set up my profile" and paste your resume. That writes into data/, which the README says is gitignored, so your resume does not end up in a commit. Then paste a job posting and say "evaluate this". You should get the six-part A-F assessment with a 1.0 to 5.0 score. After that, "tailor my resume" for your top matches, and "help" at any point to see what is available.
A realistic first session looks like this: set up the profile, evaluate three postings from different companies, and compare the B sections. If the Background Match section is thin for a posting you thought was a strong fit, the profile you pasted is probably too short, not the skill broken. The README does not document a profile schema, so the resume you paste is the profile.
Where career-ops breaks down or is the wrong tool
The privacy claim has a boundary the README states plainly, and it is worth reading twice. Data stays local, but the scan skill makes ATS API calls and the compensation section triggers web searches for salary data. If your search is confidential, for example you are employed and looking, the scan skill is the part that talks to company systems. The README does not describe what those ATS calls send or whether they are authenticated with your credentials. That is a gap you should resolve by reading the skill before pointing it at an employer's portal.
The apply skill is the second limitation. The README lists it as filling out application forms, with the prompt "Help me with this application". That wording is not accidental: the skill assists, and the README does not claim it submits applications on your own. Anyone expecting unattended mass application should read that line carefully. Form-filling is also where ATS quirks live, and the README documents no per-ATS handling.
Maintenance is the third issue, and here the repository name is the signal. It reads do-not-fork-currently-updating-v2-. The last push was on 2026-07-23, and the README carries no release notes, no version number, and no changelog. There are no releases retrieved. If you fork it now, you are forking a moving target, which is what the name warns against. The README also does not document rollback, migration, or what changes between versions, so an upgrade is an unquantified risk.
career-ops versus the original santifer/career-ops
The real alternative here is the upstream project the README credits: santifer/career-ops. The difference is the runtime, not the idea. The upstream repository is a standalone career-ops project; this one is adapted for Claude Cowork, which means the skills are packaged as a Cowork plugin with a .claude-plugin/ manifest, agents/, commands/, and skills/ directories, and you invoke them by talking to Claude rather than by running a separate application.
That distinction decides which one you want. If you already work inside Claude Cowork and want the skills available in the same conversation where you draft outreach and research companies, this plugin is the shorter path. If you want the career-ops workflow independent of Cowork, or you want to modify the evaluation logic without learning the plugin layout, the upstream project is the place to look. The README does not compare the two beyond the attribution line, so anyone choosing between them should read both repositories rather than trusting the port to be a strict superset.
The README also credits ATTRIBUTION.md for further credits, which is where the licensing relationship between the two projects is spelled out. Read it before you redistribute either one.
Licence, upgrade cost, and what the MIT terms do not cover
career-ops is MIT licensed, and the README points to ATTRIBUTION.md for credits. MIT is permissive: you can use, modify, and redistribute the plugin, including commercially, provided the copyright notice and licence text travel with it. Because this is an adaptation of another MIT project, ATTRIBUTION.md is where the upstream credit is recorded, and keeping that file intact is part of honouring the terms. This is a description of the licence, not legal advice; if you plan to ship the plugin inside a product, have someone qualified read both LICENSE and ATTRIBUTION.md.
Upgrade cost is harder to estimate because the repository publishes no releases and no changelog. The last push was on 2026-07-23. The README documents no version pinning, no migration steps, and no rollback procedure. In practice that means you should pin the commit you install rather than tracking main, especially given the repository name's warning about v2 work in progress. Your data/ directory is the part you cannot reconstruct from the repository, so back it up before you pull anything.
There is one more licence-adjacent detail: the README says data/ is excluded from git via .gitignore. That protects you from committing your resume, but it also means your profile and application history are not versioned. If you lose the directory, the plugin has nothing to work from.
Editorial conclusion
career-ops fits people running a structured search across many postings who already use Claude Cowork and are comfortable pasting a resume into a local data directory. It is the wrong tool if you want a hosted dashboard, if you need the plugin to submit applications unattended, or if you expect the repository to be stable: the repository name itself warns against forking while v2 is being updated. Before adopting it, read skills/ to confirm what each skill actually does, and check .gitignore to confirm data/ is excluded as the README claims.
Frequently asked questions
Can Claude apply to jobs for me with career-ops?
The plugin has an apply skill listed as filling out application forms, and the README's example prompt is "Help me with this application". The README does not state that the skill submits applications on your behalf, so treat it as assistance with the form rather than unattended submission.
Does career-ops send my resume or data anywhere?
The README says your data stays local and that the data/ directory is excluded from git via .gitignore. It names two exceptions where Claude reaches outside your machine: web searches for salary data and ATS API calls for job scanning.
How does the career-ops evaluation score a job posting?
You paste a posting as text or a URL and get a six-part A-F assessment covering executive summary, background match, positioning strategy, compensation and market, tailoring plan, and interview prep. The score runs from 1.0 to 5.0, and the README describes it as honest, not inflated.
Community notes