Resume Report（优化复盘 · 环节三）

Overview

The closing stage of the three-stage resume pipeline (diagnosis → optimization →
report). The deliverable is a standalone before/after comparison report page
(report.html) showing six-dimension score changes, key edits, and objective signals
measured by the script.

Artifact Handoff

Reads the outputs of stages one and two. Lookup order: user-named path →
./resume-output/ → ask the user.

• resume.html (the optimized resume) — required. If missing, ask the user for
  the optimized resume path, or suggest running stage one/two first.
• before.<ext> — the original resume (may be absent for a brand-new resume).
• diagnosis.json — the committed score baseline from stage one (may be absent).

before-score baseline (in priority order):

• If diagnosis.json is present → copy each dimension's before from its score,
  and pass --diagnosis diagnosis.json to render so the script enforces equality.
• If diagnosis.json is absent but before.* exists → judge the before scores
  honestly against the rubric and tell the user this is a degraded baseline.
• If no before.* exists → set "mode": "after_only" and omit every before field;
  do not pass --diagnosis.

Quick Routing

• User wants only diagnosis → use resume-diagnosis.
• No optimized resume yet → use resume-optimization first.
• If the target role is missing and cannot be inferred, ask one concise question.

Runtime Compatibility

• Use only relative paths from this skill folder when loading references, templates, or scripts.
• Treat agents/openai.yaml as optional Codex/OpenAI UI metadata; Claude Code can ignore it.
• Avoid relying on product-specific tools unless the user explicitly asks for that runtime's integration.
• Prefer shell commands that work in both Claude Code and Codex: python3, local scripts, and local file paths.

Required Workflow

1. Locate the three inputs above and decide the before-baseline path.
2. Read references/hr-scorecard.md and references/report-guide.md.
3. Run python3 scripts/score_resume.py measure resume.html --before before.<ext> to read objective signals. When a target role is known, add --keywords "<逗号分隔关键词>" — reuse the keyword list the diagnosis stage measured with (including any |-joined variants, so the before/after comparison stays on the same keyword list), or re-extract it from the JD / role playbook when no diagnosis exists.
4. Judge the six after scores; set before scores per the baseline path above.
5. Write ./resume-output/report.json (schema in references/report-guide.md).
6. Render: python3 scripts/score_resume.py render report.json --after resume.html --before before.<ext> -o report.html (add --diagnosis diagnosis.json when present, plus the same --keywords as step 3 so the signal table shows keyword coverage before/after).
7. Deliver: the report path and the overall score change (e.g. 2.6 → 4.3). State the before scores honestly and confirm they match the diagnosis page when a baseline was pinned.

Fairness Rules

• The before score must reflect the original resume honestly; do not deflate it to exaggerate the gain.
• Target Fit and Keyword Fit are holistic judgments; anchor them to the target role, not raw counts.
• If a dimension did not change, keep before and after equal.
• Every displayed number is script-computed; never hand-write signal counts or overall scores.

Output Shape

• Single artifact: report.html (+ optional PDF via export_resume_pdf.py).
• The overall score change and a short note per improved dimension.
• The report path.
• A closing set of options (see below) — there is no next stage.

Closing

This stage ends the pipeline. Proactively offer next steps: a variant resume for
another target role (→ resume-optimization), exporting the report to PDF, or
iterating after gathering the facts listed in the report's gaps.