Glossary
The vocabulary of an AI-powered job search, defined — ATS, the 1-5 global score, Data Contract, spray-and-pray, liveness check, tailoring, zero-token scan, STAR+R stories, and every term career-ops uses.
Every term career-ops uses, defined in one place. Terms link to the deeper reference where one exists.
ATS (Applicant Tracking System)
The software companies use to receive, store, and filter job applications — Greenhouse, Ashby, and Lever are common examples. career-ops reads their public job APIs to discover openings and can pre-fill their application forms, but never auto-submits.
Global score (1-5)
The LLM's holistic judgement of a job listing across five dimensions (match, north-star alignment, comp, cultural signals, red flags), producing a score from 1 to 5. At 4.5+ the agent recommends applying immediately; below 4.0 it recommends against applying. There is no closed-form weighting formula: the score is the LLM's judgement given the rubric, which is public on the methodology page.
Data Contract
career-ops's core architectural promise: the system layer (scripts, modes, templates) is updatable at any time, while the user layer — cv.md, config/profile.yml, data/, reports/, output/ — is never touched by an update. Your data outlives every version.
Spray-and-pray
The mass-application strategy of sending the same CV to hundreds of jobs with no targeting. career-ops explicitly rejects it: the pipeline evaluates and ranks listings first, so effort concentrates on the small set of roles worth pursuing.
Liveness check
A heuristic career-ops runs to detect whether a job listing is still open before you spend time or tokens on it. It is a best-effort signal — expired postings sometimes linger on job boards — and listings can be marked manually when the heuristic is wrong.
Tailoring
Rewriting your CV for one specific job listing — reordering emphasis, surfacing matching skills and metrics, aligning vocabulary with the job description — while keeping every claim truthful to your real experience. career-ops generates a tailored CV per application as Markdown you can edit.
Mode
A focused, prompt-defined workflow that career-ops ships as a plain Markdown file your AI CLI executes — scan, apply, tracker, interview/practice, and a dozen more. Modes are inspectable text, not black-box code.
Pipeline
The end-to-end career-ops workflow: scan portals → evaluate listings (1-5) → tailor CV → apply → track → prepare interviews. Each stage is a mode you can run independently or as a batch.
Zero-token scan
A discovery run that consumes no LLM tokens: scan.mjs reads ATS APIs, job-board feeds and career pages directly over HTTP, so finding new listings is free regardless of which AI engine you use.
AI engine
The AI coding CLI that executes career-ops prompts — Claude Code, OpenCode, Codex, GitHub Copilot CLI, and other supported CLIs — or any OpenAI-compatible endpoint. career-ops is engine-agnostic and runs on free engines too.
Question bank
A local Markdown file (interview-prep/question-bank.md) where career-ops accumulates real interview questions and your performance on each (✅/🟡/🔴). The interview/debrief mode updates it after every real interview, so preparation compounds across rounds.
STAR+R story
An interview story structured as Situation, Task, Action, Result, plus Reflection — the format career-ops uses in its story bank so behavioral answers are concrete, quantified, and reusable across interviews.
Story bank
A local Markdown file (interview-prep/story-bank.md) of your prepared STAR+R stories. Practice sessions verify answers against it, and debriefs extract new stories from what you actually said in real interviews.
Portal
A company career site backed by an ATS that career-ops can scan. Your tracked portals live in portals.yml; the scanner reads their public APIs on every run.
Batch evaluate
Scoring many saved listings in one run instead of one at a time, with flags to keep control: --limit caps the batch size, --dry-run previews what would be processed, and --resume-paused continues an interrupted run without re-spending tokens. A needs_confirmation job is an artifact-free hold that only the parent interactive session resumes after an explicit answer; retry and paused-resume flags leave it held.
Local-first
The career-ops architecture principle: everything — your data, the prompts, the scripts, the AI CLI — runs on your machine. There is no hosted backend, no account, and nothing to sign up for.
CareerOps
CareerOps is the practice of running a job search the way engineers run production: with evidence, with discipline, and with tools on the candidate's side of the table. The term names the practice, not a product; career-ops is its first reference implementation. Coined by Santiago Fernández de Valderrama Aparicio in The CareerOps Manifesto (July 14, 2026).