v2.0.2 — Template spacing sync

Syncs the blank-template layout constants to the production baseline (summary/skill after-spacing 11 to 9) and fixes the tailoring-skill pointer in Resume_Section_Spec.md. Template verified rendering exactly 1 page.

v2.0.1 — Sponsorship check documented

Documents the sponsorship check that runs before every application: H-1B history lookup (h1bdata.info / USCIS Data Hub), proven sponsors prioritized, no-history + no-signal companies skipped when better options exist. Work-authorization answers are never misrepresented.

v2.0.0 — Six-reviewer loop + interview vote

ApplyForge v2.0.0 rebuilds the per-application quality loop.

v2.0.0 — 2026-09-30

The quality loop was rebuilt from two review passes into a full six-reviewer
gauntlet with an interview vote. This is the loop that now runs before every
single application.

Added

• Review C — staff / peer engineer. Checks technical depth: would the
  tooling and scale claims survive a technical screen? Catches misused
  terminology and fluff an engineer would spot.
• Review D — executive skim. The 6-second test: title + summary + first
  two bullets must communicate who the candidate is and why they are strong
  instantly; includes a one-sentence pitch test.
• Review E — HR red-flag screen. Timeline gaps/overlaps, title-scope
  consistency, level fit vs the JD, verification risk on every claim.
• Review F — AI-voice / buzzword detector. Dedicated pass that scans the
  banned-word list plus AI phrasings (em dashes, semicolons, "not only/but
  also", triple parallelisms, "furthermore/moreover", "in today's
  fast-paced", uniform bullet rhythm, vague intensifiers without numbers,
  hedged claims like "helped with"/"assisted in") and rewrites anything
  that reads machine-generated.
• Interview vote (final gate). All six reviewers vote "interview" or "no
  interview" with a one-line reason. ALL SIX must vote interview for the
  resume to ship; any "no" triggers a targeted revision and a re-vote.
• 10/10 bar. The target is 10/10 from every reviewer. Any score below 10
  triggers a targeted revision pass and a re-score (max 3 full rounds); any
  remaining gap or dissent is shown to the user with the reason.
• Lane-specific title descriptors. The title line is now
  <JD role> • <lane descriptor> with a descriptor per lane (SRE,
  Capacity, Production Engineering, Infrastructure, Reliability,
  Performance) — no more one-size-fits-all subtitle, and never the wrong
  lane's flavor (e.g. no FinOps subtitle on an SRE resume).
• Hard impact gate (Review B). Every Experience bullet must carry
  ownership, scale, or a measurable outcome. Duty-only bullets fail
  regardless of keyword coverage. Only attested facts may satisfy the gate —
  never invented numbers.
• Summary first-line rule. Every summary opens with years of experience,
  lane identity, and the strongest attested scale or outcome for that lane.
• Skill-experience match rule. Every skill must trace to real experience
  (anchored by an Experience bullet or explicit user attestation) — no
  aspirational or JD-only skills.
• Single title per employer. One role title per Professional Experience
  entry; dual/stacked titles are banned.
• skills/jd-resume-review-loop/SKILL.md shipped in the repo. The full
  loop — all six reviews, scoring rules, interview vote, style bans — is now
  a portable skill file anyone can load into Muse or another agent.

Changed

• Quality loop expanded: Draft 1 -> ATS review -> Draft 2 -> hiring-manager review -> final is now Draft 1 -> A (ATS) -> Draft 2 -> B (hiring manager) -> Draft 3 -> C (peer) -> D (exec skim) -> E (HR) -> F (AI-voice) -> interview vote -> final.
• README and HOW_TO_RUN_MUSE rewritten around the six-reviewer loop; the
  "two review passes" language is gone everywhere.
• Version bumped to 2.0.0 (package.json).

Privacy

• No personal data in this release. The repo ships a blank template only:
  names, contacts, employers, dates, and metrics live in env vars and the
  user's private Drive, never in the repository (enforced by .gitignore:
  baseline/, jobs/, *.pdf, *.docx).

---

v1.0.0 — ApplyForge (Meta Muse edition)

First public release.

• One-prompt run on Meta Muse: base resume in, 50–75 tailored applications out per day
• JD-tailored 1-page ATS-optimized resumes (ATS + hiring-manager review loop)
• First 2 applications need approval, then full autopilot
• Google Drive filing: PDF + DOCX + JD + every question/answer, applied-only
• Job sourcing: 60 min – 7 days fresh, sponsorship-first, LinkedIn first