DEV Community

Ajit Singh
Ajit Singh

Posted on

I Built 26 AI Career Tools for Claude Code — Here's What I Learned

#career #opensource

Job hunting is a second full-time job. And if you're a developer, it's a particularly strange one — because you spend your days building tools that make other people's work easier, and then you go home and manually tailor your resume for the fifteenth time this month.

I got tired of it. So I built something.

placed-skills is a collection of 26 AI career tools that live inside Claude Code and Cursor. Resume builder, ATS checker, interview coach, salary negotiation, job tracker — all in your terminal, all accessible through natural language.

The Problem

The existing tools aren't bad. But they all have the same problem: they live in the browser. When you're in the middle of a job search, you're constantly context-switching. Five context switches for one task. Multiply that by every resume tweak, every interview prep session, every application you want to track.

I wanted something that lived where I already was.

Why Claude Code Skills (and MCP)

Claude Code has a skills system. You put a SKILL.md file in ~/.claude/skills/<skill-name>/, and Claude activates it automatically when you describe relevant tasks. No slash commands. No configuration. It just works.

The MCP angle came later. By packaging the same tools as an MCP server, I could reach Cursor users, Cline users, Continue.dev users, Zed users, Windsurf users — anyone with an MCP-compatible client.

What's Inside

26 tools across 5 categories:

Resume Builder (12 tools): Create and edit resumes with AI, 37 professional templates with ATS scores, export as PDF/DOCX/Markdown

ATS Optimizer (7 tools): Real ATS compatibility scoring (not generic tips), keyword gap analysis vs. specific job descriptions, match score with apply/don't-apply recommendation

Interview Coach (8 tools): Company-specific mock interviews, 13 system design cases, behavioral question banks with STAR format guidance, answer bank

Career Tools (12 tools): Cover letter generation, salary data by role + location, LinkedIn profile generation, salary negotiation scripts, offer analysis

Job Tracker (5 tools): Full pipeline (WISHLIST → APPLIED → INTERVIEWING → OFFER → REJECTED), conversion analytics

How to Install

Claude Code:

/plugin marketplace add Exidian-Tech/placed-skills
Enter fullscreen mode Exit fullscreen mode

Cursor / MCP clients — add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "placed": {
      "command": "npx",
      "args": ["-y", "@exidian/placed-mcp"],
      "env": { "PLACED_API_KEY": "your-key" }
    }
  }
}
Enter fullscreen mode Exit fullscreen mode

Get a free API key at https://placed.exidian.tech

What I Learned

The skills format is powerful but the ecosystem is small. ClawHub is early. MCP is winning in terms of adoption. I'd probably lead with MCP next time and treat skills as a bonus.

The job tracker was the most requested feature. People want to track applications without opening Notion or a spreadsheet.

System design cases are the most used interview prep feature. 13 cases isn't enough — I'm adding more.

Links

MIT licensed. PRs welcome. Happy to answer questions in the comments.

Top comments (1)

Collapse
 
max_quimby profile image
Max Quimby •

Interesting take on the skills vs MCP ecosystem. Agree that MCP is winning on adoption right now, but I think that's partly because the skills format is underappreciated — it's way simpler to author (just markdown + scripts) and doesn't need a running server.

Your observation about the job tracker being the most requested feature makes total sense. People don't want another SaaS for tracking — they want it where they already are. The terminal-first approach removes a huge friction point.

Curious about a few things:

  1. How are you handling the ATS scoring? Is it rule-based (keyword matching, format checking) or are you using the LLM to assess compatibility? The difference matters a lot for accuracy.

  2. For the salary data — what's the source? Levels.fyi API, BLS data, or LLM knowledge? This is one area where hallucination risk is real and users might make financial decisions based on the output.

  3. Have you thought about making the job tracker data portable? Being locked into one tool's local storage is the same problem you're solving for the browser tools.

Nice project — the "AI tools that live where developers already work" thesis is clearly where things are heading.