Claude skill · architecture & information flow

Career Strategist

A Claude Code skill that acts as a career coach, employment advisor, and negotiation rep — available 24/7, remembers everything, costs nothing.

The problem, the skill, the why

ProblemHigh-stakes
multi-threaded

A senior job search is a high-stakes operation: dozens of roles, each needing a tailored resume, cover letter, interview prep, and comp strategy. A human career coach costs thousands and isn’t available at 11pm before a morning interview.

What it doesKnows you,
remembers you

Ask any career question. The skill knows you — your career arc, wins and losses, strengths and blind spots, the jobs that worked and the ones that didn’t. It reads your vault for context, applies tested coaching methods, and gives honest, specific answers. It writes back after every conversation, so nothing is forgotten.

Why it mattersCompounding
knowledge

The difference between a good outcome and a bad one isn’t effort — it’s the right information at the right moment, remembered across conversations, applied by someone who knows your blind spots and won’t let you paper over them. Every interview, rejection, offer, and insight feeds back, so you’re never starting from zero.

Pipeline overview

Six stages from question to learning. The triage determines which mode executes; file loading is tiered by relevance.

1Career
question

Entry — a career question arrives.

2Classify
mode

Triage — route the question to one of seven branches.

3Load vault
files

Tiered by relevance — Tier 0 always, Tier 1 for a target, Tiers 2–3 on request.

4Execute

Run in one of the three operating modes — Coach, Advisor, or HR Rep.

5Deliverable

Assessment, prep, offer review, or scripts.

6Write
learnings

Feedback loop — append the session to learnings.md.

Triage treeseven branches

Every career question routes through one of seven branches. Each triggers specific vault files, activates a mode, and produces a structured output.

BranchCompany /
Role

Assess role fit→Role Assessment Format + Red Flag screen

Prep for interview→Hiring-manager one-pager, 5 questions, 5 stories, landmines map

Build resume / cover letter→Tailored .docx + cover letter + credibility artifact

Research company / person→Deep research (web + Glassdoor + LinkedIn)

Evaluate offer→Offer Evaluation + Decision Framework + Premortem

BranchStrategy /
Narrative

Shape my story→Career Coach mode (GROW) with pattern recognition

Where to focus next?→Win/Loss analysis + coaching session

Am I on track?→Evidence-bank check + pattern reflection

BranchNegotiation /
Comp

How much to ask for?→Comp benchmarks + market data by role/stage

Respond to offer/counter→Negotiation script (8 scenarios, exact words)

Equity grant fair?→Equity benchmarks by company stage

Review offer letter→6-point review: comp, benefits, red flags, negotiation, background check, recommendation

This vs. that?→Decision matrix + scenario planning (best / most likely / worst)

BranchTimeline /
Urgency

Hold out or take it?→Timeline recommendation (Phase 1–4 framework)

What patterns show?→Pattern-synthesis memo (5 questions, brief format)

BranchEmotional /
Resilience

I got rejected→Rejection Recovery Protocol (5 steps)

I feel like a fraud→Imposter-type ID + evidence-bank redirect

Nothing’s working→Pattern interrupt: shift activity, narrow focus, check the story

BranchNetworking /
Pipeline

Reach out to someone→Informational-interview script (4-part, 15 min)

Need more opportunities→Tier activation + 5-3-2 LinkedIn rhythm

Update my pipeline→Read JOB_TRACKER.md, reconcile with learnings.md

BranchSelf-
protection

How to protect myself?→Documentation plan (Receipts, Burn, Blueprint folders)

Disclosure question→Prepared script + legal-strategy guide

Three operating modes

The skill shifts fluidly between modes — a single conversation may use all three. When modes conflict, the tension is surfaced explicitly.

Career CoachPattern
recognition

Honest narrative work.

Names the pattern before offering options. The throughline: translating technical capability into organizational adoption.

GROW coaching model · pattern-synthesis memos · imposter-type identification · evidence-bank technique.

GROWCareer Construction TheoryMercenary MindsetRejection Recovery

Employment AdvisorExecution
engine

Deep research, tailored applications.

Research on companies and roles; tailored applications, interview prep, credibility artifacts, market intelligence.

Role assessment (Strong / Solid / Partial / Stretch) · stakes → action → outcome stories · hiring-manager one-pagers · red-flag detection.

STAR replacement5-question bankRed Flag Screen5-3-2 LinkedIn Rhythm

HR / Negotiation RepProtective
skeptic

Catches what you miss in the excitement of a close.

Compensation analysis, offer evaluation, equity assessment, scope protection.

Offer evaluation with market context · equity benchmarks by stage · negotiation scripts for 8 scenarios · decision framework with premortem.

Comp BenchmarksDecision MatrixScenario PlanningOffer Letter Review

File loading protocol

The agent reads vault files in prioritized tiers. Tier 0 loads on every conversation; Tier 1 loads when working on a specific company or role; Tiers 2 and 3 load on explicit request.

Tier 0 — always loadHigh-signal
context

Small, high-signal files read on every conversation — they establish the honest narrative, cross-session memory, and current state: career-background.md, learnings.md, and the context-and-preferences file.

Tier 1 — company / roleTarget-specific
intel

Loaded when working on a specific company or role: the JD file + existing resume/cover letter for that company, interview transcripts if they exist, hiring-manager research from web search, and the offer letter / terms for evaluation.

Tier 2 — application summariesDeep project
descriptions

Six application summaries covering major projects — load the one matching the role orientation: agent evaluation platform, enterprise strategy platform, enterprise search agent deployment, contract review automation, voice AI interview platform, learning & enablement platform.

Tier 3 — full contextOn explicit
request

Loaded only when needed: master-resume.md (every experience claim), application-tracker.csv (pipeline check), the legal-strategy guide, and JOB_TRACKER.md (historical win/loss analysis).

Mercenary mindset

Every job has an expiration date. Loyalty is to the mission, not the logo. Four disciplines turn the philosophy into daily practice.

1 · Know
the goal

Define success in measurable KPIs before starting — “deploy 3 production agents serving 200+ users within 90 days.” If the organization can’t articulate the mission, that’s the first problem to solve.

2 · Get the
work done

Ship, then ship again. Build credibility through output, not presence. The best defense against reorgs and budget cuts is being the person whose work everyone can point to.

3 · Surface blockers
early

Don’t suffer in silence. Name the obstacle, propose the fix, ask for what you need. If the organization won’t unblock you, that’s a data point, not a failure.

4 · Be so good they
can’t ignore you

Build things that matter, document the impact, make the work undeniable. When budgets get cut, demonstrated ROI survives; when nobody knows what you did, you’re a line item.

Feedback loop

After every significant event — interview, rejection, offer, negotiation, insight — the agent writes a dated entry to learnings.md. This file IS the agent’s memory; without it, every conversation starts from zero.

Learnings entry format

## YYYY-MM-DD — [Event Type]: [Company or Topic]
**What happened:** [1–2 sentences]
**Learning:** [1 sentence]
**Action:** [what changes based on this]

Pattern synthesis runs every 2 weeks: what’s repeating, what’s improving, what’s getting worse, and the single highest-leverage action right now.