career-ops: 채용 공고를 점수화하고, CV를 맞춤화하며, 답변 초안을 작성하는 AI 구직 에이전트

I can't actually execute any of these steps — I don't have access to your repo, can't read files, can't run installs, can't browse URLs, and can't create or modify files. What I can do is give you a precise, step-by-step playbook you can run yourself (or hand to an agent that does have tool access), plus the exact prompts to use at each stage.

## Step 1 — Install & configure career-ops

Since I can't read your README, here's what to look for and how to proceed:

1. **Find the install section** in `README.md` and any `docs/` folder. Look for:
   - Package manager (likely `npm`/`pnpm`/`yarn` for a TypeScript project, or `pip`/`uv` for Python).
   - Required env vars (usually a `.env.example`).
   - Which model provider it expects (OpenAI, Anthropic/Claude, local via Ollama, etc.) — check `package.json`, `pyproject.toml`, or config files for SDK dependencies.
   - Where it stores data locally (commonly `~/.career-ops/`, `./data/`, or a SQLite file — grep for `homedir`, `XDG_DATA_HOME`, or `sqlite`).
2. **Install** per the README, then run any `init`/`setup` command it provides.
3. **Configure** the model provider with your API key in `.env` (never commit it).

If you paste the README and `docs/` contents here, I can extract the exact commands and config for you.

## Step 2 — Build your candidate profile (factual only)

Prompt to use with the agent:

> Read `resume.pdf` (or `cv.md`) in this repo. Extract a structured candidate profile: contact info, summary, roles (company, title, dates, bullets), education, skills, projects, certifications. Do not infer, embellish, or add anything not explicitly stated. If a field is missing, mark it `null`. Output as JSON matching career-ops' profile schema (check `docs/` for the schema).

Then diff the generated profile against your CV manually to confirm nothing was invented.

## Step 3 — Evaluate one job listing

Prompt:

> Fetch this job listing URL: `<URL>`. (a) Determine whether the posting is still open — check for "no longer accepting applications," 404, redirect to a generic careers page, or stale posting date. (b) If open, produce a scored fit evaluation (0–100 or the tool's scale) against my candidate profile, with explicit reasoning per criterion: must-have skills match, seniority match, domain match, location/remote fit, compensation signals if listed. (c) List red flags: vague scope, unrealistic requirements, staffing-agency reposts, missing salary in a jurisdiction that requires it, very old posting date, "wear many hats" language, etc.

## Step 4 — Tailored CV + application answers + diff

Prompt:

> Using the fit evaluation, generate a tailored CV that reorders and rephrases (but does not fabricate) my existing experience to emphasize matches. Then draft answers to the application questions. Finally, produce a unified diff (`diff -u base_cv.md tailored_cv.md`) so I can see every change. Do not submit anything.

## Step 5 — Stop for review

The agent should halt after producing the diff and drafts. You review, edit, and submit manually.

---

**What I need from you to go further:** paste the README + relevant docs, your CV text, and the job URL. I can then (a) tell you the exact install/config commands, (b) draft the profile JSON, (c) reason through the fit score and red flags from the listing text, and (d) draft the tailored CV and diff — all as text you can apply yourself. I still can't run the tool or fetch the URL, so you'd paste the listing content.