career-ops

AI job search command center — evaluates job offers with A-F scoring, generates ATS-optimized CV PDFs, scans multiple job portals in parallel, tracks application pipelines, and analyzes rejection patt

SKILL.md

GitHub

career-ops -- AI Job Search Command Center

Skill for The Agency. Delegates to the most relevant specialized agent based on the mode requested. For any career-ops task, invoke the right sub-agent rather than handling it directly.

Mode → Agent Routing

Mode Agent Description
scan Portal scan (general-purpose + /career-ops) Scrape portals + on-demand: node scan-positions.mjs --position "AI Engineer" --location hanoi
scan:pos scan-positions.mjs On-demand scan for a specific position title + location. Usage: npm run scan -- --position "Digital Marketing Manager" --location hanoi
oferta / ofertas Offer Evaluator A-F scoring, archetype detection, gap analysis, comparison/ranking
pipeline Pipeline Strategist Process inbox URLs: evaluate → report → PDF → tracker (never touch applications.md directly)
batch Batch Processing Lead Parallel worker pipeline for bulk evaluations
patterns Pattern Analysis Specialist Rejection pattern detection, actionable recommendations from tracker + reports
pdf CV Specialist ATS-optimized CV generation with Playwright HTML→PDF
tracker Pipeline Analyst Application status overview, funnel metrics
deep Discovery Coach Deep company research
apply Outbound Strategist Live application form assistant with Playwright
contacto Outbound Strategist LinkedIn outreach: find contacts + draft messages
training Trend Researcher Evaluate course/cert against career goals
project Game Designer Evaluate portfolio project idea
interview-prep Discovery Coach Company-specific interview prep with STAR+R story bank

Routing Logic

  1. Parse {{mode}} from the invocation args
  2. If empty/no args → show the discovery menu inline (no sub-agent needed)
  3. If JD text or URL → auto-route to Pipeline Strategist (pipeline mode)
  4. Otherwise → spawn the matching specialized agent with the mode as the task description
  5. Spawn the matching specialized agent from the Mode → Agent Routing table above.
    • Use the agent name/type from column 2 of the table
    • Pass prompt: "mode: {mode}, project: {current_project}"
    • Never use general-purpose here — every career-ops mode has a named agent

Discovery Menu (no args)

career-ops -- Command Center

Available commands:
  scan        → Discover new offers across portals (3-level scanner)
  oferta      → Evaluate a job offer (A-F scoring, archetype detection)
  ofertas     → Compare and rank multiple offers
  pipeline    → Process pending inbox URLs (evaluate → report → PDF → tracker)
  batch       → Bulk evaluations via parallel workers
  patterns    → Analyze rejection patterns from tracker + reports
  pdf         → Generate ATS-optimized CV PDF
  tracker     → Application funnel and status overview
  apply       → Live application form assistant
  contacto    → LinkedIn outreach (find contacts + draft message)
  deep        → Deep company research
  interview-prep → Company-specific interview prep (STAR+R stories)
  training    → Evaluate course/cert against your career goals
  project     → Evaluate a portfolio project idea

Paste a job description or URL → runs the full pipeline automatically.

Sub-Agent Prompt Template

When spawning a specialized agent:

You are running in career-ops mode: {MODE}

Your task: {MODE_DESCRIPTION}

Project context:
- career-ops root: {current_project_dir}
- CV: cv.md (read this first)
- Profile: config/profile.yml
- Tracker: data/applications.md
- Pipeline inbox: data/pipeline.md

Key rules:
- NEVER submit applications on the user's behalf
- Scores below 4.0/5 → recommend against applying
- NEVER edit applications.md directly — write TSV to batch/tracker-additions/
- After every batch: run node merge-tracker.mjs
- All reports must include **URL:** in header
- Use Playwright (not WebSearch) to verify offers are active

Execute the {MODE} mode fully. Report results back to the parent session.

Global Conventions (all agents)

  • Ethical gate: Always discourage scores < 4.0/5. Quality > quantity.
  • User review before submit: Draft everything, STOP before clicking Submit.
  • TSV pipeline additions: Write to batch/tracker-additions/{num}-{slug}.tsv, never touch data/applications.md directly.
  • Report numbering: Sequential 3-digit zero-padded, max existing + 1.
  • Verification: Always use Playwright to verify offer is still active before evaluating.
  • Canonical states: Evaluated | Applied | Responded | Interview | Offer | Rejected | Discarded | SKIP — no bold in status field.