career-ops

oferta

Runs the A–H evaluation against one listing and records it in your tracker, without generating a PDF. Use it when you only need the scoring decision and not the full pipeline.

What it does

oferta mode reads a single JD (URL or pasted text), scores it against the rubric (five dimensions plus a holistic global score) with cited evidence, and writes the evaluation report. No PDF, but it records the listing in your tracker with the status Evaluated; evaluating the same company and role again updates that row instead of adding a second one. Fast and cheap.

When to use it

Use oferta to triage a borderline listing quickly. If the score clears 4.0 and you decide to apply, escalate to auto-pipeline or apply mode.

What a run looks like

You hand the mode one listing and it returns the evaluation, with a score and cited reasoning: Blocks A through F, a Block G legitimacy read and a one-line verdict, plus Block H draft answers when the score is 4.5 or higher. Paste JD text directly, or pass a URL and the mode first confirms the posting is still live so a dead link never burns a full evaluation.

Terminal
/career-ops oferta https://boards.greenhouse.io/acme/jobs/4021
Evaluation excerpt
# Evaluation: Acme — Staff AI Engineer

**Score:** 4.3/5
**Archetype:** LLMOps
**Legitimacy:** High Confidence

## A) Role Summary
| Field      | Value  |
|------------|--------|
| Domain     | LLMOps |
| Seniority  | Staff  |
| Remote     | Full   |

## G) Posting Legitimacy
Assessment: High Confidence — posted 3 days ago, apply button active,
JD names specific tools and first-6-month scope.

## Verdict (lead)
**Apply.** Strong match on the role's core axis with no hard stops,
and Block G returned High Confidence.

Blocks B through F cover CV match and gaps, level strategy, comp and demand, a customization plan, and STAR+R interview stories mapped to the job's requirements.

The verdict comes right after Block G: the call in bold (Apply, Consider, Research first or Skip) and one reason drawn from the blocks above. It restates the decision recorded in the report's machine summary and never adds a new judgment, so a reader who stops there still leaves with the answer.

When to reach for it

Reach for oferta to score a single listing quickly without committing to the full pipeline. Worth knowing:

  • Pass a URL and a liveness check runs first — an expired or 404 posting is stopped before Block A, so you never spend a run on phantom content.

  • A LinkedIn page whose description stays a loading skeleton counts as JD unavailable, not as a closed job: oferta stops before Block A and asks you to paste the job text or give an employer or ATS link for the same role.

  • The archetype is classified against your own target roles, the Your Target Roles section of modes/_profile.md; the default role types apply only when you have defined none. A role that matches none of your targets scores 1 on North Star alignment, the evaluation continues on the other dimensions, and the report's Archetype line reads Not a target — closest default: {default role type} or Not a target — no close match.

  • Company and comp research is bounded to at most five web searches; this is an evaluation, not deep research. Run /career-ops deep separately when you want the fuller company picture.

  • Block G returns one of three tiers — High Confidence, Proceed with Caution, or Suspicious — so likely ghost jobs surface before you invest interview effort.

  • When the role reached you through a recruiter (a LinkedIn message, an email or a call) instead of a posting you found yourself, Block G also checks the recruiter. Five checks: whether the company's own careers page lists the role, whether an unfamiliar name is attached to a known brand, whether the recruiter writes from a domain the company controls, public scam reports, and requests to run code they send, share your screen, move to Telegram, WhatsApp or Discord, or pay for anything. Recruiter contact on its own counts as neutral. Two concerning checks, one of them a careers page without the role or a scam report, make the tier Suspicious and add a recruiter identity note. career-ops never clones or runs code a recruiter sent.

  • When the posting explicitly describes a fixed-term role, such as an 18-month contract or a temporary position, Block G adds a fixed-term contract note that quotes the phrase, plus an optional talking point for the compensation conversation. It is informational only: it never changes the score, the legitimacy tier or the recommendation.

  • In a CLI that lets the agent rename its own session, the session takes the posting's name, such as VP Infrastructure, Acme Health, once the job description is read. Set session_title: off in config/profile.yml to keep your own session titles.

Score Evidence

The report includes Score Evidence after Risk Summary. It covers five dimensions: CV match, North Star alignment, compensation, cultural signals, and red flags. Each row records an evidence status (supported, partial, or unknown), a concrete source or observation, and an unresolved question. The section ends with an evidence-confidence tier (High, Medium, or Low), its reason, and up to three verification priorities.

This tier describes the evidence behind the Global Score; it is neither a hiring probability nor Block G's posting-legitimacy tier, and it does not change the numeric score. Machine Summary preserves the five statuses in score_evidence, the tier in confidence, and priorities in confidence_gaps. When there are no priorities left, confidence_gaps is an empty list.

On this page