Resume customizer

Resume Customizer — Senior PM Toolkit

BEFORE BUILDING, MIRROR YOUR GENERIC RESUME. Once you have one, your maintained generic resume (a real .docx you keep current — see STEP 0 below) is the authoritative design benchmark. Build every tailored resume by mirroring that file — render it and diff against it — NOT by copying a prior company's build script. A copied build script is a styling starting point, never the spec; it silently drifts from your current standard.

The one conflict rule, split by domain (these two clauses do not overlap):

• FORMAT / layout / voice / structure — your maintained generic resume is the benchmark of record. It is a real .docx you keep current, so on any disagreement about how the resume looks or reads, the generic wins over this skill's templates. Templates teach style; the canonical is the truth about form.
• A contested factual METRIC (a number that differs between the generic, YOUR_PROFILE.md, and this skill) — do NOT silently pick the generic, or any copy. Re-derive the figure from the most primary source (your own evidence / the verified-metrics table), fix every copy in one pass, and flag the conflict to the user. A resume can carry a stale number; the primary source settles it.

QUICK REFERENCE

Core function: Generate ATS-optimized resumes and cover letters for director-level PM roles.

Reference files (use when needed):

• references/star-stories.md — your STAR stories (success + failure)
• references/interview-prep.md — RCA, guesstimate, case study, questions to ask
• references/career-tools.md — outreach templates, salary scripts, LinkedIn optimization, 90-day plans
• references/detailed-context.md — full work history, certifications, side projects, strengths
• references/evaluation-framework.md — 7-block JD scoring before generation
• references/story-bank-index.md — cross-indexed STAR story selection
• references/github_in_resume_pattern.md — GitHub introduction pattern
• references/referral_email_pattern.md — forwardable referral email pattern

Profile source: ${CLAUDE_PLUGIN_ROOT}/references/YOUR_PROFILE.md is the single source of truth for the user's background.

---

STEP 0 — START FROM CANONICAL ARTIFACT (BLOCKING)

Before generating ANY resume, find and read the user's canonical generic resume on disk. Customize via minimal targeted edits to that canonical. Do not rebuild from this skill's templates.

The skill's templates and detailed-context.md teach STYLE and RULES. The canonical artifact on disk is the TRUTH about content — current end dates, availability tags, header links, latest side projects, exact metric values, partner-specific phrasing already validated with real referrers. Templates drift; the canonical does not.

Canonical paths (checked in priority order — the default needs no setup beyond dropping your file in):

1. ${CLAUDE_PLUGIN_ROOT}/references/generic-resume.docx — the default canonical. Drop your maintained generic resume here (it is git-ignored). This fixed path is read first, so a fresh user is never blocked on an unfilled placeholder.
2. ${CLAUDE_PLUGIN_ROOT}/references/generic-resume.pdf — PDF sibling at the same location, for a fast read via pypdf.
3. {{path/to/your/generic-resume.docx}} — OPTIONAL override, only if you keep your canonical somewhere else. Leave the placeholder as-is to use the default path above. ⚠️ A custom path must live OUTSIDE this repository (e.g. ~/Documents/resumes/). If you insist on a custom location inside the repo, add that exact path to .gitignore in the same step — only the default paths above and output/ are ignored for you; an arbitrary in-repo path is committable.
4. {{path/to/prior-applications/}}<company>/ — most recent shipped customization for a similar role type. ⚠️ Same rule: keep this folder OUTSIDE the repository, or add it to .gitignore — prior applications are complete personal documents.
5. If files have moved: find ${CLAUDE_PLUGIN_ROOT}/references {{your resume folder}} -iname "*[Rr]esume*[Gg]eneric*" -o -iname "generic-resume.*" (where {{your resume folder}} is a location outside this repo, per the rule above)

Procedure:

1. Open the canonical (.docx via python-docx or .pdf via pypdf). READ IT IN FULL, do not skim.
2. If the default canonical (path 1) is missing or out of date: check path 2, then the optional override (path 3), then path 4, then the find in path 5. If ALL fail, STOP and ask the user: "Canonical resume not found at references/generic-resume.docx or the other paths. Where is the current one, or should I generate the first one now?" Only then fall back to template-based generation from detailed-context.md, with an explicit warning: "WARNING: no canonical found, building from the template summary. This will likely miss recent items like updated end dates, new side projects, and exact metric values. Proceed?" This is the one legitimate template-first case (a fresh install with no canonical yet). When the user approves that first resume, SAVE the approved .docx to ${CLAUDE_PLUGIN_ROOT}/references/generic-resume.docx — it becomes the canonical every later run reads.
3. Copy the canonical to the target filename in the output directory (${CLAUDE_PLUGIN_ROOT}/output/{{LastName}}_{{FirstName}}_Resume_{{Company}}.docx).
4. Apply minimal targeted edits for the JD. Typical edits:
   • Promote one bullet to lead position if the JD demands it (e.g., the activation bullet first for a Growth PM role).
   • Sharpen 1-2 bullet prefixes for JD-specific keywords.
   • Swap 2-3 Expertise keywords for JD-specific terms.
   • Adjust the Summary anchor line if positioning needs a tilt (rarely the whole Summary).
5. After edits, diff against the canonical (unzip -p both .docx files' word/document.xml and diff). A diff surfaces what changed — it flags unexpected drops so you can eyeball them; it does NOT by itself prove the resume is semantically complete or that layout and hyperlinks survived.
6. Render the edited .docx to PDF and look at it (e.g. LibreOffice soffice --headless --convert-to pdf, or open it). Confirm visually: 2 pages, every section present and in order, LinkedIn/portfolio hyperlinks still clickable, no layout break. This render-and-look step is what actually catches a dropped section or a broken link; the XML diff only points you at candidates.
7. Verify ATS-clean (0 em dashes, 0 arrows, 0 currency symbols, 0 curly quotes).

Why this is BLOCKING: Skipping this step causes systematic loss of validated content that a summary file cannot recover — most-recent end dates, availability tags, header links, latest side projects, exact metric values, phrasing already battle-tested with referrers. Build-from-context introduces drift; copy-and-edit preserves what is already validated.

---

MISTAKES TO AVOID

Section structure — permanent removals

• No Strategic Focus section. It duplicates Summary positioning and bores the reader. Pillars, themes, focus areas — all redundant when the Summary already anchors the narrative.
• No Key Achievements section. It duplicates Experience bullets.
• No standalone "OEM & Telco Partnerships" section. Fold partnership content into the role bullets where it was earned.
• No standalone "Earlier Career" section. Fold early roles directly into Professional Experience with one compact bullet each.

STAR framework — implicit, never labeled

• NEVER write (S), (T), (A), (R) prefixes on resume bullets. That reads kiddish and undermines senior positioning. STAR is an internal checklist for writing bullets, not scaffolding to display.
• Weave Situation, Task, Action, Result into natural narrative prose. A good bullet opens with context, flows through action, and lands on measurable results — the reader gets STAR structure without seeing the framework.
• Bullet format: bold theme prefix + flowing narrative. 40-55 words typical for senior roles with depth; 12-20 words for early career single-line bullets.

Bad (kiddish, never do this):

Market entry (STAR): (S) Company needed entry... (T) Build distribution... (A) Ran user research... (R) 60X growth...

Good (natural, senior-level):

Market entry from scratch: Entered a new region when monetization was {Nx} the baseline market but no playbook existed. Ran user research, built a {strategy descriptor} strategy, navigated relevant compliance, and led product integration for distribution partners post-contract. Scaled {start size} to {end size} ({growth multiplier}) in {timeframe}; key metric improved {X to Y}; retention improved by {Z}.

Content and claims

• Do NOT invent metrics not in YOUR_PROFILE.md's verified tables.
• Do NOT claim experiences the user doesn't have.
• Do NOT exaggerate scope or impact.
• Do NOT use any metric the user has explicitly marked as "REMOVED" in YOUR_PROFILE.md.
• Match the user's accuracy guardrails from YOUR_PROFILE.md's "Never say" / "Always say" lists.
• If the user joined an existing platform, do NOT claim "scaling from 0 to N" — clarify their actual contribution.
• If the user is one of several peers (e.g., one GPM of many on a roadmap), use collaborative framing ("Collaborated with the broader product team to...") rather than ownership claims ("I own...").

Always verify

• Every metric must come from the verified tables in YOUR_PROFILE.md.
• Every achievement must trace to actual experience.
• If uncertain, use conservative framing or omit.
• Be specific about what the user actually owns (team, markets, product areas).

---

LANGUAGE RULES

Avoid vague language

Never use:

• "Lead the platform" — lead WHAT specifically?
• "Responsible for" — what did you ACHIEVE?
• "Worked on" — what was YOUR contribution?
• "Helped with" — junior language

Use instead:

• "Own [specific area] including [scope]"
• "Drive [metric] through [action]"
• "Built [thing] that achieved [result]"
• "Manage [team size] across [geographies]"

Avoid passive/weak language (cover letters)

Never use:

• "I am exploring opportunities"
• "I would welcome the chance"
• "I'd welcome a conversation about..." (FORBIDDEN — passive)
• "Please feel free to reach out"
• "I am open to"
• "I believe I would be a good fit"
• "X years of experience" — years don't matter, impact does

Use instead:

• "I own..." / "I drive..." / "I built..."
• "Let's connect — {{phone}}." (confident close with contact)
• "Happy to connect: {{phone}}" (action-oriented)
• "My direct wins include..." (ownership language)

Avoid AI-sounding language

Never use: pivotal moment, key turning point, evolving landscape, serves as a testament, underscores the importance, comprehensive, robust, streamlined, fostering, cultivating, showcasing, at its core, in order to, "Additionally" at sentence start, em dashes for emphasis.

Write like a human: simple constructions (is/are/has), natural sentence length variation, specific over inflated, have opinions where appropriate, acknowledge uncertainty when relevant.

MANDATORY: Use Humanizer Skill

Before outputting ANY resume or cover letter, Claude MUST:

1. Generate the document using this skill's templates.
2. Read ${CLAUDE_PLUGIN_ROOT}/skills/humanizer/SKILL.md.
3. Apply humanizer checklist to all generated text.
4. Remove any AI-sounding patterns that slipped through.

---

ATS COMPLIANCE (MANDATORY)

Character replacements

Replace	With	Example
Currency symbols (₹, €)	ISO code ("INR", "EUR", "USD")	"USD 5M ARR" or "EUR 12M GMV"
→ (arrow)	"to"	"$10 to $15"
0→1	"0-to-1"	Hyphenated
— (em-dash)	"-"	Standard hyphen only

Output format — Golden Standard

• File Type: .docx (primary), PDF if requested
• Font: Calibri (not Arial)
• Filename: {{LastName}}_{{FirstName}}_Resume_{{Company}}.docx
• Save location: the git-ignored ${CLAUDE_PLUGIN_ROOT}/output/ directory — always. Generated resumes and cover letters hold real personal data and must never land in a tracked path. The cover letter uses the same directory: output/{{LastName}}_{{FirstName}}_CoverLetter_{{Company}}.docx.
• Hyperlinks: LinkedIn and portfolio must be clickable

Exact styling (match these values for golden-standard output):

• Name: size 28 (14pt), bold, color #2E5090
• Contact line: size 18 (9pt), color #555555
• Section headings: size 22 (11pt), bold, color #2E5090, bottom border (single, sz=4, color #2E5090, space=1)
• Body text: size 20 (10pt), color #1A1A1A
• Sub-descriptions: size 20, italic, color #555555
• Role headers: size 20, title bold #1A1A1A, company normal, dates right-aligned via tab stop pos=9026

Page layout: US Letter (12240 x 15840 DXA); margins top=620, right=720, bottom=620, left=720; line spacing 276 (1.15), lineRule=auto.

Bullet style: Word ListParagraph style with numPr numbering (or dash - equivalent). Bullet character: dash. Indent: left=720, hanging=360. Spacing: before=20, after=20.

What to avoid

• Tables, columns, text boxes
• Graphics, images, icons
• Headers/footers with content
• Special characters (use ASCII equivalents)
• Multiple font styles

Date format

Use: "January 2023 - Present". Not: "Jan 2023", not "01/2023".

2-page limit (HARD)

If exceeding 2 pages, cut in this order:

1. Trim current-role bullets to 5 max (from 6).
2. Drop second-tier achievement bullets.
3. Condense previous-role bullets.
4. Shorten side-project descriptions.
5. Trim Expertise line length.

---

PER-CLAIM DEFENSIBILITY AUDIT (MANDATORY before "final")

The rule

Before declaring any resume / email / cover letter "final" or "95% confident," build a per-claim audit table. Each claim → source → confidence percentage. Surface anything below 90% to user before they ship.

Audit table format

Claim	Source	Confidence	Risk
[each verbatim claim from draft]	[profile metric, STAR story #, public source, or USER INPUT]	[%]	[LOW / MEDIUM / HIGH]

Three categories of miss to specifically watch for

1. Math errors

• Always verify percent calculations against the underlying numbers. AND specify the metric type when a percent could be ambiguous. Example: "$10 to $15 (+25% YoY)" — clear that the percent is annualized growth, not the math result of the lifetime $5 increase. Without "YoY," readers will calculate $10→$15 lifetime as +50% and the parenthetical reads as a math error.
• Always verify multiplier claims. Arithmetic: "{start} to N×{start}" is N times, not whatever sounds bigger. Example: "20% to 80%" is 4X (not 60% better — that's a different operation). When numbers are inherited from prior resumes, RE-VERIFY against YOUR_PROFILE.md. Do not trust.
• When numbers are inherited from prior resumes, RE-VERIFY. Do not trust.
• Keep the unit label attached to every figure (devices vs users vs DAU vs sessions vs revenue). Unit confusion is how a wrong metric becomes self-reinforcing: a "6M" that silently changes units survives weeks of reuse because every copy looks internally consistent.
• A verification rule can itself be wrong. If a metric's guard-rule in this skill and your canonical resume / YOUR_PROFILE.md disagree, re-derive the number from the primary source — never let a metric's own guard become the source of truth. A wrong rule enforces the wrong number on every future generation until someone checks it against the primary.

2. Scope overclaims — verbs that overstate role

Wrong (overclaim)	Right (honest)	Why
"Closed [partner] distribution"	"Led product integration for [partner] post-contract"	BD closes contracts; product integrates after MoU
"Architected the SDK"	"Framed the architecture problem and drove the design with engineering"	Product gives problem statement; engineering designs
"I own X feature/roadmap"	"Collaborated with the broader product team to ship..."	Match YOUR_PROFILE.md's explicit ownership boundaries
"Scaled X from 0 to N"	"Led GTM and 0-to-1 launches for [specific market/segment]"	If user joined an existing platform
"The same pattern as [X]"	"A similar [structural element] to [X]" / "The kind of [X] that..."	Identity claim → analogy claim. Defensible if probed.

3. Inherited awkward phrasings — phrases lifted from rules/templates that read odd in context

• "Our [team type] team" / "broader [team type] team" → use specific org name from profile
• "Tight latency budget" without numbers → drop or anchor with a real number
• Buzzword taglines as keywords ("Distributed Team Leadership") → reads as filler unless backed by specifics
• Bare labels like "AI/ML" → fine for cert blocks; reads thin in email body. Use "AI building" or specific subfield.

Process

1. Draft the content
2. Run humanizer skill pass (em dashes, copula avoidance, AI tics)
3. Build the per-claim audit table
4. For each claim below 90% confidence: propose softening or dropping, surface to user
5. Only after the user resolves the flagged items, declare "ready"
6. Do NOT say "95% confidence" without having built the table

Referrer-feedback corollary

If a referrer is reviewing the artifacts before forwarding, expect 4-7 iterations. Each round is more data about what the recruiter will read. Don't lock after v1. See referral_email_pattern.md for the full iteration pattern.

---

PRE-GENERATION CHECKLIST

Before generating ANY resume or cover letter:

Step 0: Locate Canonical Artifact (MANDATORY FIRST STEP)

• Read the canonical at ${CLAUDE_PLUGIN_ROOT}/references/generic-resume.docx (the default path; or its .pdf sibling, or your optional override path)
• If it is missing or unclear, run find ${CLAUDE_PLUGIN_ROOT}/references {{your resume folder}} -iname "*[Rr]esume*[Gg]eneric*" -o -iname "generic-resume.*" ({{your resume folder}} = a location OUTSIDE this repo — see STEP 0's custom-path rule)
• READ the canonical in full via python-docx or pypdf — do NOT skim, do NOT rebuild from detailed-context.md memory
• If none exists (fresh install), generate from templates ONCE, and on approval SAVE it to ${CLAUDE_PLUGIN_ROOT}/references/generic-resume.docx so it becomes the canonical
• Copy the canonical to the target filename in the git-ignored output directory: ${CLAUDE_PLUGIN_ROOT}/output/{{LastName}}_{{FirstName}}_Resume_{{Company}}.docx
• All subsequent steps apply TARGETED EDITS to the canonical, not rebuilds from templates
• After final edits: diff against the canonical (unzip -p both word/document.xml and diff) to SURFACE unexpected drops, THEN render the .docx to PDF and visually confirm 2 pages, section order, and clickable hyperlinks (the diff alone does not prove completeness or layout)
• See the STEP 0 section above for the full procedure + canonical path hierarchy (and the one legitimate no-canonical-yet exception)

Step 1: JD Analysis

• Company name and role title
• Company type: AI / Consumer / Enterprise / Fintech / Telco / Other
• Top 3 must-haves
• Top 3 nice-to-haves
• 8-10 keywords to mirror
• Honest fit score (1-10)

Step 2: Resume Structure — LOCKED ORDER

1. Summary (3-4 line narrative anchor; lead with strongest credential)
2. Expertise (ATS keyword row — placed HERE for maximum ATS scoring, before Experience)
3. Professional Experience (ALL roles including early career folded in)
4. AI Product Building / Side Projects (ALWAYS after Experience, even for AI companies — credibility first)
5. Education / Certifications / Recognition

Critical rules:

• NO Strategic Focus. NO Key Achievements. NO standalone partnership section. NO standalone Earlier Career section.
• Section heading: "PROFESSIONAL EXPERIENCE" (not "Career Experience", not "Experience").
• Early career roles fold INTO Professional Experience with roleHeader + one compact bullet (12-20 words each). No depth, no STAR treatment.
• Side projects / AI building always after Professional Experience (credibility first, even for AI companies).

Header:

• Name from YOUR_PROFILE.md contact section
• Location (City, Country)
• Email (clickable)
• Phone
• LinkedIn (clickable)
• Portfolio (clickable, if any)
• GitHub (only when relevant to the role family — see github_in_resume_pattern.md)

Summary (3-4 lines) — Narrative Anchor:

• Open with strongest credential (degree, certification, or current title — choose what hooks fastest)
• Describe HOW the user works (methodology, not just what)
• Minimal metrics — scope references are fine; specific numbers go in Experience
• Sounds like a human, not a list of facts

Expertise Section (after Summary, before Professional Experience):

• Single line or two lines, pipe-separated
• Mirror exact JD keywords where user has real experience
• Include: skills, methodologies, tools, domain terms from JD

Professional Experience bullets — mandatory elements:

• Product description (italic subheading) under each company name
• Bold theme prefix + natural narrative body
• PM craft signals (discovery, roadmap, stakeholder management)
• Decision context (WHY you did it, not just WHAT)
• Measurable result with verified metric from YOUR_PROFILE.md
• STAR elements WOVEN — never labeled
• Each bullet 40-55 words for senior roles; 12-20 words for early career

Side Projects / AI Product Building (if relevant):

• Portfolio link first (if any)
• List specific projects with one-line proof points
• Tech stack categorized

Education:

• Pull from YOUR_PROFILE.md education section
• Honors / awards if any

Recognition (if included):

• Aggregate awards with prestige framing (consecutive-year patterns, top-X language)
• Do NOT list cute internal award names that read odd externally
• One or two named awards max if multi-year, with the prestige descriptor

Step 3: Cover Letter

• Company name used (not "your organization")
• Explicit product category match if relevant
• Strongest credential in opening paragraph
• Specific ownership language
• Research hook (company news/initiative/product)
• NO passive phrases (see Language Rules)
• Strong confident close WITH phone number ("Let's connect — {{phone}}.")
• 250-300 words max

Step 4: Metrics Verification

Pull from YOUR_PROFILE.md's verified metrics tables. If a metric is NOT in those tables, DO NOT USE IT.

Step 5: ATS Compliance

• No special characters
• LinkedIn and Portfolio are clickable hyperlinks
• Font: Calibri
• Dates: "January 2023 - Present" format
• Filename: {{LastName}}_{{FirstName}}_Resume_{{Company}}.docx, saved into the git-ignored ${CLAUDE_PLUGIN_ROOT}/output/ directory
• Section heading: "PROFESSIONAL EXPERIENCE"

Step 6: Page Check

• Exactly 2 pages (not 1, not 3)

Step 7: Company-Type Checks

For AI Companies:

• AI / side-project section after Experience (not before)
• Portfolio link prominent
• Shipped-AI achievements in Experience bullets
• AI certifications included

For Consumer Scale:

• Headline scale metric prominent in sub-description
• Best-growth bullet as lead
• Multi-market scope emphasized (if applicable)

For Enterprise / B2B:

• Partnership / Fortune 500 work folded into bullets (NOT standalone section)
• Regulatory expertise mentioned (if applicable)
• B2B2C model explained where relevant

For Fintech:

• Payment / GMV work prominent
• Bank / processor partnerships mentioned
• Compliance language (PCI, RBI, etc.) if applicable

For Telcos:

• Telco partnerships visible
• B2B2C distribution model explained
• Explicit product category match in cover letter

Step 8: Humanizer Check (MANDATORY)

• Read ${CLAUDE_PLUGIN_ROOT}/skills/humanizer/SKILL.md
• Apply humanizer checklist to ALL generated text
• Sentence lengths vary naturally
• Specific over inflated language

Step 9: Skills Form Mapping (for application forms)

When filling an application form's skills field:

• Extract exact keywords from JD first
• Verify EACH keyword against the user's actual experience
• Include only skills with real, demonstrable experience
• Flag marginal skills and ask before including
• Never add skills that appear in JD but NOT in profile
• Format: comma-separated list

---

RESUME OUTPUT SPECIFICATION

Header Format

{{FULL NAME}}
{{City}}, {{Country}} | {{email}} | {{phone}}
LinkedIn: {{linkedin URL}} | Portfolio: {{portfolio URL}}

Both LinkedIn and Portfolio must be clickable hyperlinks.

Summary Template (3-4 lines — Narrative Anchor)

Purpose: Establish WHO the user is and HOW they work. Lead with strongest credential. Minimal metrics.

Structure: Credential + current title + approach/methodology + scope hint.

Default (Platform Leader anchor):

{{Positioning tagline — 3-5 words pipe-separated}}

{{Credential}} and {{Current Title}} at {{Company}}. {{2-3 sentences describing
HOW you work — methodology, not what you do. Use first person. Read it aloud:
does it sound like a human or a press release?}}
{{One scope-hint sentence — team size, market count, headline metric.}}

The reference files have company-type-specific summary variants. Adapt the language to match the JD's tone (more technical for AI, more business for fintech).

Expertise Section (immediately after Summary)

Purpose: ATS keyword row. Recruiters and ATS systems scan skills first. Format: Single line or two lines max, pipe-separated.

EXPERTISE
{{Keyword 1}} | {{Keyword 2}} | {{Keyword 3}} | {{Keyword 4}} | {{Keyword 5}} |
{{Keyword 6}} | {{Keyword 7}} | {{Keyword 8}} | {{Keyword 9}}

Customize per JD: Swap in exact keywords from the job description where the user has real experience. Mirror the JD's exact phrasing where possible.

Professional Experience — Bullet Structure

Format: Natural narrative prose with bold theme prefix. STAR woven implicitly — never labeled.

Section heading: "PROFESSIONAL EXPERIENCE"

Bullet length guidance:

• Senior roles: 40-55 words; STAR depth implicit
• Early career: single compact line, 12-20 words

Example structure (replace bracketed content with user's actual experience):

{{Current Title}} | {{Current Company}} | {{Start month/year}} - Present
{{One-line italic product description — what the product is, key scale metric}}

- {{Theme prefix in bold}}: {{Context that explains why this work mattered.
  Action you took with method (not just outcome). Result with verified metric
  from profile.}} 40-55 words.

- {{Theme prefix}}: {{Same structure — context, action with method, result.}}

[3-5 bullets per senior role]

Early Career — Single Compact Bullet

Format: roleHeader (title, company, month+year range) + one single-line bullet (12-20 words). NO depth, NO STAR treatment.

{{Title}} | {{Company}} | {{Date range}}
- {{One-line achievement with verified metric.}}

Side Projects / AI Product Building Section

Always after Professional Experience.

AI PRODUCT BUILDING (or SIDE PROJECTS — pick what matches user's portfolio)
{{One-line framing about why these projects exist — keep technical edge, ship daily, etc.}}

- {{Project name}}: {{Problem solved + tech stack hint + concrete result}}
- {{Project name}}: {{Same structure}}
- {{Project name}}: {{Same structure}}

Tech: {{Stack categorized — LLMs | automation | data | integrations}}

Education

EDUCATION
{{Institution}} | {{Degree}} | {{Years}} | {{Honors if any}}
{{Institution}} | {{Degree}} | {{Years}} | {{Location}}

Certifications (5 max, categorized)

CERTIFICATIONS
{{Category 1}}: {{Cert 1}} | {{Cert 2}} | {{Cert 3}}
{{Category 2}}: {{Cert 4}} | {{Cert 5}}

Recognition (aggregate with prestige framing)

Always lead with the pattern (consecutive years, top-X) and name the prestigious-sounding award specifically. Do NOT list cute internal award names that read odd externally.

RECOGNITION
{{Pattern statement — e.g., "Company top performer recognition X consecutive years
({{years}}), including the {{prestigious award name}} {{N}} times ({{years}}) —
the company's highest individual contribution award."}}
{{Other recognition — alumni boards, external awards}}

---

COVER LETTER SPECIFICATION

Format

• Length: 250-300 words
• Tone: Confident, specific
• Structure: Narrative arc (not checklist)

Narrative Arc (preferred — scores 9/10 vs 7/10 for checklist)

• Para 1 — WHY this move: Honest motivation for why this company makes sense. Open with a specific product/market observation; include strongest credential naturally.
• Para 2 — What you bring: Connected story with method. Research first, strategy, iterate on data. Weave credential naturally (if not in para 1).
• Para 3 — Specific JD mapping: 2 proof points told with depth, not 3 points told thin. Show the journey (diagnosed X, did Y, result Z).
• Para 4 — Direct close: Phone number required. No passive language.

Strong openers

• "I build products in markets where there are zero users and zero playbook."
• "I don't just strategize about AI — I ship it."
• "I build the same type of product you're building."
• "When [COMPANY NEWS], I recognized a challenge I've solved before."

Strong closes (MUST include phone)

• "Let's connect — {{phone}}."
• "Call me on {{channel}}: {{phone}}."
• "Happy to discuss in detail: {{phone}}."

FORBIDDEN closes

• "I'd welcome a conversation about..." (passive — FORBIDDEN)
• "I am exploring opportunities..."
• "Please feel free to reach out..."
• "Happy to discuss..." (without phone number)
• "Looking forward to discussing how I can contribute..." (generic)

Template

[Date]

Dear {{Hiring Manager / Hiring Team}},

{{PARA 1 — WHY this move, 3-4 sentences. Specific product/market observation.
Include strongest credential.}}

{{PARA 2 — What you bring, 3-4 sentences. Method, not list.}}

{{PARA 3 — Two proof points with depth. Situation → action → result.}}

Let's connect — {{phone}}.

{{Full Name}}
{{phone}} | {{email}}
{{LinkedIn}} | {{Portfolio}}

Generic Cover Letter (for network forwarding)

Dear Friends and Network,

I'm exploring my next role and would appreciate your help. After {{N years}} at
{{Current Company}} {{brief scope description}}, I'm seeking senior product roles.

What I'm looking for:
- {{Role types — e.g., "Senior PM / Director of Product / Principal PM / Head of Product"}}
- {{Company categories — e.g., "AI-first companies, Fortune 500, or GCCs"}}
- {{Domains — e.g., "Consumer platforms, B2B2C, or Fintech"}}

What I bring:
- {{Headline win 1 with verified metric}}
- {{Headline win 2}}
- {{Headline win 3}}
- {{Headline win 4}}

If you know of relevant opportunities or people I should talk to, I'd be grateful
for an introduction. Feel free to forward this note.

Best,
{{First name}}
{{phone}} | {{email}} | {{LinkedIn}}

---

COMPANY-SPECIFIC STRATEGIES

For each archetype: what to lead with, where to place the side-projects section, and the interview angle to rehearse.

For AI Companies (Anthropic, OpenAI, Google AI, etc.)

Lead with: Shipped-AI achievements in Experience, side AI projects, AI certifications, hands-on building philosophy.

Resume order: AI / Side Projects after Experience (credibility first). Shipped-AI achievements prominent in Experience bullets. Portfolio link in header.

Interview angle: Speak to the gap between AI capabilities and real product value. Show how you've crossed it personally.

For Consumer Scale (Google, Meta, Netflix, Flipkart, etc.)

Lead with: Headline scale metric, growth multipliers, multi-market scope (if applicable).

Interview angle: Both 0-to-1 (greenfield market entry) AND 1-N (scale-up) experience. Specific examples of moving needle on a platform that's already live.

For Enterprise / B2B (Atlassian, Salesforce, Microsoft, etc.)

Lead with: Fortune 500 / enterprise partnerships, regulatory expertise (if applicable), B2B2C platform model.

Resume order: Partnership work folded into Experience bullets (no standalone section). Emphasize partnership complexity, stakeholder management, compliance expertise.

Interview angle: Distribution thinking (selling through partners to reach end users), enterprise procurement, ability to land partner-defined outcomes.

For Fintech (Stripe, Razorpay, PhonePe, etc.)

Lead with: GMV / TPV / payments work, payment gateway integrations, settlement / reliability metrics, bank partnerships.

Interview angle: Building trust in a financial context (especially if the user built fintech inside a non-fintech company — that's harder).

For Telcos (Deutsche Telekom, Verizon, T-Mobile, SoftBank, Vodafone, etc.)

Lead with: Direct telco partnerships, OEM + Telco integration experience, B2B2C distribution model, engagement / lock-screen / connectivity products.

Resume approach: Summary mentions OEM/telco partnerships; Experience bullets emphasize telco work; product description uses explicit category match.

For Director+ Roles

Lead with: Team building (X to N PMs), cross-functional leadership, strategic influence (any company-direction story), P&L language (GMV, ARPU, unit economics).

Interview angle: Director-level impact = shaping company direction, not just executing. Strategic insights that changed where the company invested.

---

SOURCE OF VERIFIED METRICS

Every numeric claim in a generated resume MUST trace to a row in YOUR_PROFILE.md's verified-metrics tables. If a metric is not in those tables:

1. Do NOT use it.
2. Ask the user whether to add it (with source) or omit it from the bullet.
3. If the user adds it, update YOUR_PROFILE.md's verified-metrics table at the same time.

Every claim traces to a source. The verified-metrics table is a table of claims each traced to where it was verified — a profile row, a canonical-resume bullet, or explicit user input. Two rules keep it from rotting:

• When a metric is corrected, correct it in EVERY copy of YOUR data in the same pass — the profile's verified-metrics table, your canonical resume, and any prior drafts in output/ you intend to reuse. All of those live in git-ignored files, and that is exactly where corrections belong: NEVER write your corrected real metric into this tracked SKILL.md — its examples are fictional teaching content and stay fixed (this file is committed, so anything you write here would be published if the repo is pushed). A stale copy left behind in your data becomes the next self-reinforcing error (it will be inherited by the next resume and re-verified against itself).
• On a contested factual METRIC (the profile vs the canonical resume vs a prior draft): do not silently pick one — not even the generic. Re-derive from the most primary source available, fix all copies in your git-ignored data files, and flag the conflict to the user. (This is the metric half of the STEP 0 conflict rule; FORMAT/layout/voice disagreements resolve the other way — the generic resume wins.)

---

VERSION HISTORY

• v2.0 (Template refresh): Ported the generic mechanisms that evolved upstream: STEP 0 canonical-artifact-first generation (BLOCKING), generic-resume-as-authoritative-benchmark banner (mirror + render + diff; never copy a prior build script as spec; generic wins on conflict), and anti-self-reinforcing metric rules (unit labels stay attached; a metric's guard-rule is never its own source; corrections propagate across your git-ignored data files in one pass, never into this tracked skill).
• v1.0 (Template): Generic version extracted from a real director-PM job search workflow. All personal data replaced with placeholders. Methodology preserved verbatim.

---

END OF SKILL