Glossary

The same terms, in the same meanings, as the project's canonical engineering glossary — phrased here for humans first. Where a term has its own guide page, it's linked.

A

Application Pipeline — the end-to-end journey /apply runs: posting in → fit evaluation, tailored CV, cover letter, review, compiled PDFs out.

Application Tracker — the CSV file recording every application: company, role, status, fit rating, and links to the exact documents sent. Feeds the dashboard and the deduplication in /search.

Application Status Enum — the seven allowed values of the status column: Draft, Sent, Interview, Offer, Rejected, Withdrawn, Closed. The dashboard enforces which moves between them are legal.

Atomic CSV Write — how the dashboard saves your tracker without ever leaving a half-written file, even if something interrupts it mid-save: write a temporary copy first, keep a backup, then swap it into place in one step.

ATS-Safe Export — the machine-readable companions /apply produces beside the polished PDF — a plain-text .txt and a Word .docx from the same content — so an applicant-tracking system can parse your CV reliably. A self-check confirms the text is recoverable from the PDF.

Allowed-Tools — an internal safety list declaring which tools each CareerForge skill may use; anything not listed is refused.

B

Behavioral Profile — the part of your profile describing how you work: style, strengths, growth areas, preferred environment. Shapes the cover letter's voice and the culture-fit score.

Business Rules Engine — the fixed scoring frameworks and validation rules behind fit evaluation, content cutting, and the final quality checks.

C

Candidate Profile — the structured record of your identity, education, experience, and skills. Built by /setup; the source of truth every CV claim must trace to.

Competency Expansion — discovering skills your documents imply but your profile doesn't list yet (planned: the /expand command).

Compile-and-Inspect Loop — after drafting, documents are compiled to PDF and visually checked — page counts, orphaned headings, font mismatches — and re-fixed until clean.

Content Cutting (Relevance-Weighted) — when a CV runs over two pages, the lines least relevant to this posting are cut first, wherever they sit.

Cross-Reference Check — during onboarding, your sources are compared against each other (CV vs LinkedIn vs diplomas); any disagreement is shown to you to resolve, never silently picked.

D

Deal-Breaker — a hard constraint of yours (no relocation, minimum salary) that disqualifies a job regardless of every other score.

Drafter Agent — the primary AI in the pipeline: evaluates fit, writes the documents, applies the reviewer's feedback.

Document Folder — documents/, where you drop CVs, LinkedIn exports, diplomas, and reference letters for /setup to scan.

E

Evaluation Framework — the five-dimension scoring system: Technical Skills 30%, Career Alignment 30%, Experience 25%, Behavioral Fit 15%, Location pass/fail. See /apply.

F

Fabrication Audit — the check, run by /apply, that traces every CV claim back to the profile file backing it and flags any it can't. You read the result in the Provenance panel.

Fit Assessment (Quick) — the fast High / Medium / Low ranking /search gives each new posting. Triage, not the full evaluation.

Fit Rating — the 0–100 weighted score from the full evaluation. Column 9 of your tracker; the fuel for the fit charts.

Forward-Looking Framing — the cover-letter principle: write about the problems you'll solve for them, not a recap of your CV.

G

Gap Heatmap — a prioritized table of skill gaps between your profile and target postings (planned: the /upskill command).

Ghost Job — a posting that's advertised but isn't really being filled — left open to collect résumés, gauge a market, or look like growth. One of the red flags the posting legitimacy check watches for.

I

Idempotent Operation — safe to run twice: re-running /setup with the same inputs duplicates nothing and overwrites nothing.

Interview Backtrack Test — the quality bar for every claim: could you back this sentence in an interview without "well, what I actually meant was…"? If not, it gets softened or cut.

J

Job Portal Adapter — a pluggable connector for one job board, so new boards can be added without touching the core (planned for a future milestone).

L

Language Registry — the set of UI languages CareerForge ships, currently 12 (machine-translated, in beta), with English as canonical. See Languages & Locales.

Learning Plan — a structured study guide produced from skill-gap analysis (planned: /upskill).

Legitimacy Verdict — the standalone judgment /apply gives a posting — Verified, Caution, or Suspicious — with cited reasons, kept separate from the fit score so a warning can't be averaged away.

Locale Pack — a per-market bundle of CV conventions (document term, page size and count, photo norm, personal fields, date format, GDPR clauses) that /apply applies for the market you're targeting. Separate from UI language. See Languages & Locales.

N

Named Sub-Agent — an internal mechanism for defining extra AI agents (the reviewer is one) with their own instructions and, optionally, their own model.

O

Onboarding — /setup's three convergent paths — scan documents, import a CV, or interview — all ending in the same profile.

P

Profile Files — the seven documents defining your professional identity: candidate profile, behavioral profile, writing style, evaluation framework, CV templates, cover-letter templates, interview prep.

Pipeline KPI — the dashboard's headline numbers: status counts, average fit (30 days), interview rate (90 days). See the KPIs.

Posting Legitimacy — the question of whether a job posting is genuine and safe to engage with, assessed by /apply before drafting and surfaced as a Legitimacy Verdict. The system warns; you decide.

Profile Statement — the 3–5 line pitch at the top of your CV, kept in several variants and tailored per application.

Provenance Panel — the dashboard view, in an application's detail drawer, that shows each CV claim beside the profile source backing it, with unbacked claims flagged. Read-only. See the Provenance panel.

R

Read-Only Mode — a dashboard flag that disables all editing — handy for demos, screenshots, or guaranteeing an accident-free look around.

Red Flag — a warning sign in a job posting — upfront fees, personal-data harvesting, a mismatched domain, a too-good salary — that feeds the Legitimacy Verdict.

Reviewer Agent — the second AI in /apply: fresh perspective, researches the company, critiques both drafts. It suggests; the drafter (and you) decide.

S

Salary Benchmark — optional salary comparison against data you supply — union statistics, surveys, your own research (the salary_lookup tool).

Search Queries — your configured search phrases, organized by priority, that /search runs across your chosen portals.

Seen Jobs Registry — the local file remembering every posting you've ever been shown, so /search never repeats itself.

SKILL.md Orchestrator — the internal anchor file each CareerForge skill ships with, declaring what it does and which tools it may use.

STAR Format — Situation, Task, Action, Result: the interview-answer structure. /setup builds STAR stories from your real experience and leaves clearly marked stubs where only your memory can fill the details.

T

Tracking Dashboard — the local web UI over your tracker file. Runs only on your machine (127.0.0.1), edits only status and notes, and renders everything else read-only.

Two-Plane Skill Architecture — the internal split between knowledge-style skills (instructions the AI reads) and tool-style skills (external programs it runs).

Traceability — the project's practice of linking every requirement to the design, code, and tests that satisfy it, via IDs like REQ-… and TC-….

V

Verification Checklist — the final pass/fail list /apply runs once at the end: factual accuracy, targeting, consistency, LaTeX quality, and PDF layout.

W

Workspace Settings — the local configuration file declaring which tools the AI may use in your copy of CareerForge. Personal, and never committed.