Cv Customizer

---
name: cv-customizer
sheetId: "7.01"
description: "5-phase CV optimization workflow — job analysis, story extraction, fit scoring, career progression analysis, and CV rewrite with Google Doc output."
category: people-talent
project: cv-customizer
platform: claude
status: Done
author: Peter Tu
input: "Job description (text, URL, or file) and candidate CV/resume"
process: "5-phase pipeline — job decoding, SCQA story extraction, fit scoring, career progression analysis, and section-by-section CV rewrite"
output: "Optimized CV as a Google Doc, with fit scores and progression report"
synergy: []
type: agent
skills: ["cv-job-decoder", "cv-story-extractor", "cv-fit-scorer", "cv-progression", "cv-rewrite"]
house-style: bound

---

# CV Customizer

```bash
npx skills add https://github.com/peter-tu-zynkr/zynkr-skill-builder --skill cv-customizer
```

Orchestrate a 5-phase CV optimization workflow: analyze the target job, extract candidate stories, score fit, analyze career progression, and rewrite the CV — then output as a Google Doc. Use this skill when applying for a specific role and you have both a job description and a CV ready to tailor.

---

## Step 1 — Collect inputs

Ask the user for:
1. **Job Description**: paste text directly, or provide a URL / file path
2. **Current CV / Resume**: paste text directly, or provide a file path

**Detect input type:**
- If pasted text -> use directly
- If file path (.txt, .md, .pdf, .docx) -> read the file
- If URL -> use WebFetch to retrieve the page content and extract the JD

Store both inputs as `JD_TEXT` and `CV_TEXT` for use across all phases.

---

## Step 2 — Phase 1: Unpack What the Company Wants (subagent)

Display:
```
---------------------------------------------
Phase 1: Unpack What the Company Wants
---------------------------------------------
Analyzing job description...
```

Launch the `cv-job-decoder` agent using the Agent tool:
- Pass the full `JD_TEXT` as the prompt
- Wait for completion

Display the structured output (Role Context Summary, Skills Table, KPI Mapping, ATS Keywords).

Then ask:
```
Does this analysis look accurate? (Yes / Adjust)
```

- If "Adjust": ask what to change, re-run or manually correct
- If "Yes": store the output as `PHASE1_OUTPUT`, proceed to Phase 2

---

## Step 3 — Phase 2: Contextualized Story Extraction (interactive)

Display:
```
---------------------------------------------
Phase 2: Story Extraction (SCQA)
---------------------------------------------
```

Read the instructions from:
`./agents/cv-story-extractor.md`

Execute the Phase 2 flow **in the main conversation**:
1. Compare `CV_TEXT` against `PHASE1_OUTPUT` skill priorities
2. Identify 1-3 underrepresented skills
3. For each skill, guide the user through SCQA (Situation, Complication, Question, Answer)
4. Confirm each story summary with the user
5. Collect all stories as `PHASE2_OUTPUT`

---

## Step 4 — Phase 3: Fit Scoring & Benchmarking (interactive)

Display:
```
---------------------------------------------
Phase 3: Candidate Fit Scoring
---------------------------------------------
```

Read the instructions from:
`./agents/cv-fit-scorer.md`

Execute the Phase 3 flow **in the main conversation**:
1. Score the CV across 4 dimensions (Relevance, Value, Narrative, ATS) — each 1-5
2. Present each dimension one at a time, wait for user acknowledgment
3. Show benchmark comparison table
4. Present action summary
5. Store as `PHASE3_OUTPUT`

---

## Step 5 — Phase 4: Career Progression Analysis (subagent)

Display:
```
---------------------------------------------
Phase 4: Career Skill Progression Analysis
---------------------------------------------
Analyzing career trajectory...
```

Launch the `cv-progression` agent using the Agent tool.
Pass all accumulated context in the prompt:
```
## Context from Prior Phases

### Phase 1 Output:
[PHASE1_OUTPUT]

### Phase 2 Output (SCQA Stories):
[PHASE2_OUTPUT]

### Phase 3 Output (Fit Scores):
[PHASE3_OUTPUT]

### Candidate CV:
[CV_TEXT]

---

Analyze the career progression following your 4-checkpoint chain.
```

Wait for completion. Display the full progression report.

---

## Step 6 — Phase 5: Rewrite & Final Quality Check (interactive)

Display:
```
---------------------------------------------
Phase 5: CV Rewrite & Quality Check
---------------------------------------------
```

Read the instructions from:
`./agents/cv-rewrite.md`

Execute the Phase 5 flow **in the main conversation**, using all prior phase outputs:
1. Rewrite Executive Summary -> Accept / Modify / Skip
2. Rewrite each Work Experience role -> Accept / Modify / Skip
3. Rewrite Skills section -> Accept / Modify / Skip
4. Review Education -> Accept / Modify / Skip
5. Final polish and changes summary table

Assemble the final CV content as `FINAL_CV`.

---

## Step 7 — Output to Google Doc

Display:
```
---------------------------------------------
Creating Google Doc...
---------------------------------------------
```

1. Use `mcp__google-workspace__create_doc` to create a new Google Doc:
   - `user_google_email`: `<your-google-workspace-account>`
   - Title: `CV - [Candidate Name] - [Target Role] - [Date]`

2. Use `mcp__google-workspace__batch_update_doc` to write the final CV content into the doc, following the template structure:
   - Header (name, contact info)
   - Executive Summary
   - Work Experience (each role with bullets)
   - Education
   - Skills

3. Use `mcp__google-workspace__get_drive_shareable_link` to get the doc URL.

4. Display completion summary:

```
---------------------------------------------
CV Customizer Complete
---------------------------------------------

Target: [Company] - [Role]

Phase Results:
  [1] Job Analysis      - [key insight]
  [2] Story Extraction  - [N] SCQA stories captured
  [3] Fit Score         - [N]/20 overall
  [4] Progression       - [N] priority rewrites identified
  [5] CV Rewrite        - [N] sections rewritten

Google Doc: [URL]
---------------------------------------------
```

---

## Error Handling

- If Phase 1 subagent fails, display the error and ask the user to provide more JD details
- If the user wants to skip a phase, allow it and note what context will be missing downstream
- If Google Doc creation fails, fall back to displaying the final CV as formatted text in the conversation
- Never silently swallow errors — always surface them clearly

## House style

Writing style is **not owned by this file**. The house voice lives in two Google Docs under
`[@] 寫作指南` (`12DBdFz3SK22ie9im_ThFMI7IBRXsTZsV`), read at runtime:

- 《[2.0] Zynkr 通用風格指南 House Voice》 `10bOIQwRm9Pxwgct4hlwCwK_B4Pipai1HqBPZKzyRHSE` —
  the universal core, plus the addendum for this surface
- 《[3.2] 禁用詞清單 Forbidden Words》 `1N5sHLP4qzmmhpCGsi6KElxi1z0MFe4QZ0Q_35T10Uyg`

Read both before producing client- or reader-facing text, and scan the draft against 《[3.2]》
before handing it over. If Drive is unreachable, say so in the output rather than proceeding
unchecked. Never re-implement either list inside this file.