fix: broaden SKILLS section regex to handle whitespace variants

fix: bulk history panel overlap + SKILLS section diagnostics (v2.6.3)

fix: bad escape \u crash in _sanitize_for_tectonic — bulk gen 25/25 fail

fix: root cause of placed-0-keywords — gap weave consumed placement targets (v2.6.2)

feat: add refresh button + re-inject content script before bulk extraction

fix: keyword placement actually places keywords + LaTeX compile fix (v2.6.1)

feat: competency-aware keyword placement, Skills-first design (v2.6.0)

feat: V3 bulk generation from LinkedIn search pages (v2.0.0)

feat: sweep PM_SKILL_TAXONOMY for comprehensive keyword extraction

feat: extract more PM keywords, strip legal boilerplate, robust insertion

fix: placed keywords show as supported in extension + robust insertion

chore: deploy all fixes to HF, bump extension v1.12.3

fix: API crash on None estimate + all keywords earn ATS score

fix: placed keywords now earn ATS score (re-calibrate after placement)

chore: bump extension version to v1.12.2

feat: natural keyword sentence placement across resume sections

feat: expand confirmed ATS skills

feat: weave JD phrases into bullets; use Settings cover letter

fix: Skills Role Keywords row breaking resume Tectonic PDF

fix: LaTeX PDF compile on HF and honest extension compile UX

fix: flat ATS scores and zero supported phrases in extension runs

docs: record full-suite result for the score fix (3 failed / 125 passed, no regressions)

docs: backfill the missing v1.11.0 HISTORY entry + log the NIM 404

docs: record the before/after score symmetry invariant (README + HISTORY)

fix: ATS score could DECREASE after tailoring (66 -> 41 with zero edits)

feat: unified Generate Application flow with cover letter (extension v1.11.0)

perf: V1 deterministic by default; LLM opt-in via V1_ENABLE_LLM

test: deployment + API-route smoke scripts (deterministic PDF, two-parser)

merge: deterministic LLM-free V1 pipeline

feat: deterministic LLM-free ATS pipeline (extraction, rewrite, scoring)

chore: untrack unrelated Everjobs folder (binary junk, HF-incompatible)

fix: fair symmetric stuffing penalty + more extraction retries

fix: reliable live extraction (nemotron) + fair PDF-based before/after

feat: live V1 acceptance — independent audit, 2-parser PDF, summary opt, completeness

feat: V1 NIM fallback + 90% gate + PDF-scored alignment + correction pass

feat: V1 evidence-backed optimization — rewriting, calibration, scoring

fix: evidence-gated ATS pipeline — stop scraped-page contamination & fabrication

feat: LLM keyword extraction via Calibrated Keyword Match Framework

fix: V1 places ALL keywords in exp+project bullets, no Core Competencies

feat: inject 2 keyword items per project list-end before Core Competencies

fix: Core Competencies trusted-only + duplicate word-boundary check

fix: SSE stream missing scores/downloads + source name mismatch

fix: friendly status display for BELOW_TARGET_REPAIRABLE

fix: natural keyword placement + fix comma-split + relax V1 frequency filter

fix: always allow PDF download for BELOW_TARGET_REPAIRABLE

feat: auto re-extract JD on SPA navigation (job click without page reload)

feat: migrate injected panel to Chrome Side Panel API

fix: use body zoom to reflow page instead of html margin/width hack

fix: _shiftPage width not max-width + clip body overflow

fix: V1 keyword quality gate, layout fix, progress anim; add V1+V2 e2e test