[sarthakagrawal927]

Why

Tailored resumes need targeted, reversible edits and inspectable evidence behind job-fit judgments. This adds those workflow ideas while preserving approved achievement wording.

What changed

• Refine the summary or reorder one bullet/project group, review a proposal, apply or undo it, and save through existing History.
• Link job requirements to exact master-resume excerpts, bind reports to the analyzed source, and reanalyze existing scores.
• Reject unsupported claims, invalid orders and stale proposals. Keep guest storage and authenticated billing/refunds on their existing paths.

Scope

Limited to the current tailoring workspace and fit reports. No new dependencies, schema migrations or production configuration changes.

Blast radius

Touches draft generation, fit-score presentation and their review/save integration. Achievement text, identity, education and untargeted sections remain preserved. Source excerpts establish provenance, not semantic proof of qualification.

Verification

• pnpm quality: 814 tests passed, one opt-in test skipped; all lint, type, documentation and code-health gates passed without baseline changes.
• pnpm cf:build: full Next/OpenNext/Astro pipeline passed, including 18/18 parity routes.
• Guest browser checks exercised real server actions against a local synthetic AI provider at 390, 768 and 1440 px. Preview/apply/undo, exact save/reload, summary refinement, invalid-response recovery and source-checked fit evidence passed. Independent Claude review through CodeVetter covered the full change and subsequent fixes.

Production and live managed-AI behavior were not exercised. No deployment was performed.