A Claude Code plugin for schema-driven, fact-traceable resume and cover-letter automation. Every bullet on every resume traces back to a verified atomic fact, and a multi-layer validation pipeline enforces correctness, decorum, and zero fabrication.

What It Does

You capture career facts once (/careerops:capturing-fact), stored inside experience files
You drop a job description in inbox/ and run /careerops:analyzing-jd
You run /careerops:generating-resume <jd-id> and the system:
- Scores all your facts against the JD and ranks them by relevance
- Asks you to approve the presentation plan (which roles to expand, compress, merge)
- Composes bullets with full provenance (every bullet traces to a fact ID)
- Renders to PDF via RenderCV/Typst
- Runs an 8-gate validator (page budget, fabrication check, em-dash scan, etc.)
- Audits semantically for AI markers and quality

Result: a tailored, ATS-clean PDF resume with a claim ledger showing exactly which facts back which bullets.

Installation

npx careerops install

This copies the plugin to ~/.claude/plugins/careerops/ and registers it in your Claude Code settings. Restart Claude Code afterward.

Requirements:

Python 3.10+
pip install pyyaml
pip install rendercv (for PDF output)

All CareerOps skills are then available under the /careerops: namespace.

Quick Start (New User)

# 1. Create an empty career data directory and cd into it
mkdir -p ~/my-career
cd ~/my-career

# 2. Launch Claude
claude

# 3. Inside Claude, run the first-time setup wizard
/careerops:setting-up

# 4. Drop an existing resume into raw_data/ and seed your knowledge base
/careerops:seeding-career-db raw_data/your-resume.tex

Run /careerops:getting-help full inside Claude for the complete command reference.

Commands

Setup

Command	Purpose
`/careerops:setting-up`	First-run wizard: scaffold directories, collect profile, write config
`/careerops:seeding-career-db <path>`	Import a resume (.tex or .md) into the career knowledge base

Capture

Command	Purpose
`/careerops:capturing-fact`	Interview to record a new career achievement
`/careerops:capturing-evidence <fact-id>`	Attach evidence to an existing fact

Apply

Command	Purpose
`/careerops:analyzing-jd <path>`	Analyze a JD from `inbox/` — scores all facts by relevance
`/careerops:generating-resume <jd-id>`	Full pipeline: plan approval, compose, render, validate, audit
`/careerops:auditing-resume <app-id>`	Re-run the semantic auditor
`/careerops:humanizing-resume <app-id>`	Manual AI-marker cleanup
`/careerops:logging-outcome <app-id>`	Record interview/reject/offer/no-response

Health

Command	Purpose
`/careerops:linting-career`	Schema + reference + em-dash check on all career YAML
`/careerops:getting-help`	Quick-start guide. Add `full` for complete command reference

Architecture

CareerOps separates plugin (this repo: how to do things) from user data (your directory: facts, applications, profile). The plugin is portable, publishable, and contains zero personal information. Your data stays where you choose to put it.

Data model: Facts are embedded inside experience files (career/experiences/X-*.yaml) rather than stored as individual files. Every JD analysis scores all your facts by relevance, giving the composer a ranked shortlist while keeping all facts available for selection.

See:

CLAUDE.md for plugin development context, hard rules, conventions
docs/USER-SETUP.md for the new-user bootstrap guide
docs/superpowers/specs/ for design history

Hard Rules

Every output from this plugin is validated against:

No em-dashes. U+2014 and variants forbidden everywhere. Multi-layer enforcement.
No fabrication. Every resume bullet traces to a verified fact ID.
Date and employer immutability. Schema-locked fields.
Tier 2 reframing requires explicit per-application opt-in.
Encapsulation requires user approval before render.

Tech Stack

Renderer: RenderCV (Typst)
Data: YAML files with JSON Schema validation; facts embedded in experience envelopes
JD analysis: Pure Claude subagent with relevance scoring (no ESCO/SkillNER)
Validation: Python validators (8 gates) + Claude semantic auditor
Bootstrap: using-careerops skill injected at every SessionStart via hook