LinkedIn: Lead Health Check — LinkedIn Outreach Platform Agentic Workflow

schema_version: 1
version: 0.2.0
last_verified: '2026-08-25'
id: linkedin_lead_health
title: 'LinkedIn: Lead Health Check'
description: >-
  Opens a LinkedIn profile, scrolls to trigger lazy-loaded sections (about,
  experience, education, skills, activity), extracts the full profile, and
  asks the LLM to evaluate the lead against the current SideButton ICP
  (2026-08 outreach strategy: non-technical queue owners, agent teams).
  Returns structured JSON with raw profile data, detected signals, ICP
  evaluation (score/tier), and suggested outreach angle + opener. Designed
  for batch enrichment of Max's 1st-degree connections.
overview: |
  The heavyweight enrichment pass. Scrolls the full profile (about, experience,
  education, skills, activity), extracts everything, and scores the person against
  the baked-in ICP with a 0–11 rubric (title fit, company size, geography, recency,
  warmth) into tiers A–D, plus detected signals and a suggested outreach angle.

  The ICP defaults reflect the 2026-08 outreach strategy: the buyer is the owner of
  a queue (CEO/GF/COO/non-technical founder at 20–500 person software-enabled
  companies), product leaders are the operator path, agencies and fractional execs
  are channel, strong-eng-org CTOs get an advisor ask only. Pass `icp_context` to
  override without editing the pack.

  For a single person right before writing a message, `linkedin_profile_brief` is
  the faster, lighter alternative; this one is meant for scoring lists.
category:
  level: task
  domain: sales
  reusable: true
params:
  profile_url:
    type: string
    description: Full LinkedIn profile URL (e.g. https://www.linkedin.com/in/username/)
    required: true
  icp_context:
    type: string
    description: Optional override for ICP definition. Defaults to the 2026-08 SideButton outreach ICP baked into the prompt.
    required: false
  max_scrolls:
    type: string
    description: Number of scroll steps to trigger lazy-load. Default 6.
    required: false
policies:
  allowed_domains:
    - '*.linkedin.com'
steps:
  - type: browser.navigate
    url: '{{profile_url}}'
  - type: browser.wait
    selector: main
    timeout: 12000
  - type: browser.wait
    selector: main a[href*="/in/"], main a[href*="/company/"]
    timeout: 10000

  - type: browser.scroll
    direction: down
    amount: 1200
  - type: browser.wait
    selector: main
    timeout: 1000
  - type: browser.scroll
    direction: down
    amount: 1200
  - type: browser.wait
    selector: main
    timeout: 1000
  - type: browser.scroll
    direction: down
    amount: 1200
  - type: browser.wait
    selector: main
    timeout: 1000
  - type: browser.scroll
    direction: down
    amount: 1200
  - type: browser.wait
    selector: main
    timeout: 1000
  - type: browser.scroll
    direction: down
    amount: 1200
  - type: browser.wait
    selector: main
    timeout: 1000
  - type: browser.scroll
    direction: down
    amount: 1200
  - type: browser.wait
    selector: main
    timeout: 1000

  - type: browser.extract
    selector: main
    as: profile_blob

  - type: llm.generate
    prompt: >
      You are evaluating a LinkedIn profile as a potential SideButton lead.


      ## SideButton ICP (2026-08 outreach strategy)

      {{icp_context}}

      Thesis: we sell the first agent team; the buyer is the owner of a queue,
      not the owner of a codebase. Primary buyer: CEO / Geschäftsführer / COO /
      CIO / non-technical founder at a 20–500 person software-enabled company
      whose engineering is weak, absent, or fully booked on the core product.
      Operator persona (second entry path, budget usually elsewhere): Head of
      Product / CPO / PM who can describe work precisely.

      Channel (not buyer, still valuable): agencies, dev shops, IT-Systemhäuser,
      consultancies, fractional/interim execs — white-label angle. Multipliers:
      investors, analysts, coaches, community and event owners — reach and
      intros, advisor-style ask only.

      Deprioritize as a pitch: CTOs / VPs Eng / staff engineers of strong
      engineering organisations (five nos on record), anyone whose only angle
      would be "make your developers faster", recruiters, competitors.

      Geography: DACH-first, UK/NL/Nordics secondary.

      Proof assets: public Felix Ohswald post — a product lead closed 456 Jira
      tickets in a month, about 7 engineers of output (GoStudent, PAST TENSE
      only, never present-tense); 92.5% first-pass on senior review (technical
      readers only). Never lead with parallelism, VM counts, 24/7, or
      orchestration.

      Commercials: Agent Launch 10–15k EUR one-off (design partners 5k or
      credited); monthly agent teams Team ~2.5k / Fleet ~6k / Scale from 12k.


      ## Scoring rubric (total 0–11)

      - title_fit (0–3): 3 = CEO/GF/COO/owner of a 20–500 person non-eng-led
        company (queue owner); 2.5 = non-technical or commercial founder;
        2 = Head of Product / CPO / senior PM (operator); 1.5 = agency,
        consultancy, or fractional exec (channel); 1 = founder-CTO with a tiny
        or fully booked team; 0.5 = CTO/VP Eng of a strong engineering org
        (advisor ask only); 0 = unrelated/irrelevant
      - size_fit (0–3): 3 = 20–500 emp; 2 = 500–1000; 1.5 = <20; 1 = 1000+;
        0 = solo/unclear
      - geo (0–2): 2 = DACH; 1.5 = UK/NL/Nordics; 1 = US; 0.5 = other; 0 = unclear
      - recency (0–1): 1 = active/posting within 90 days; 0.8 = within 1 year;
        0.5 = dormant/no activity visible
      - warmth (0–2): 2 = strong mutual signal (shared employer, evident collaboration,
        prior thread); 1.5 = many mutual connections or shared industry events;
        1 = some mutuals; 0.5 = weak/none visible

      Tiers: A = 9–11, B = 6–8, C = 4–5, D < 4.


      ## Raw profile data

      Profile URL: {{profile_url}}


      Main content (newline-delimited; line 1 = name, then pronouns, headline,
      location, connections, About, Activity, Experience, Education, Skills,
      Languages, Interests. Parse this to fill the schema below.):

      ---

      {{profile_blob}}

      ---


      ## Your task

      Respond with RAW JSON ONLY — no prose before, no markdown fences, no
      comments. Match this exact schema (leave a field null if data is not
      visible on the profile rather than guessing):


      {
        "profile_url": "{{profile_url}}",
        "name": "string",
        "headline": "string",
        "location": "string | null",
        "connection_degree": "1st | 2nd | 3rd | out-of-network | unknown",
        "mutual_connections_count": number | null,
        "followers": "string | null",
        "has_open_to_work_banner": boolean,
        "has_provides_services": boolean,
        "has_hiring_banner": boolean,
        "about_summary": "first 500 chars of about section, or null",
        "current_role": {
          "title": "string",
          "company": "string",
          "employment_type": "string | null",
          "start_date": "YYYY-MM | null",
          "duration": "e.g. 1 yr 3 mos | null",
          "location": "string | null",
          "description_snippet": "first 200 chars or null"
        },
        "prior_roles": [
          {"title": "string", "company": "string", "start": "YYYY-MM | null", "end": "YYYY-MM | null", "duration": "string | null"}
        ],
        "prior_roles_count": number,
        "total_years_experience": number | null,
        "career_arc_signal": "repeat-CTO | first-time-CTO | IC-to-leader | founder | specialist | unclear",
        "education": [
          {"institution": "string", "degree": "string | null", "field": "string | null", "end": "YYYY | null"}
        ],
        "top_skills": ["string", "..."],
        "recent_activity": {
          "posted_within_90d": boolean,
          "recent_topics": ["string", "..."],
          "engagement_level": "active | sporadic | dormant | unknown"
        },
        "languages": ["string", "..."],
        "detected_signals": {
          "likely_dach": boolean,
          "likely_scaleup": boolean,
          "likely_buyer": boolean,
          "likely_influencer_only": boolean,
          "just_joined_role_90d": boolean,
          "posting_about_ai": boolean,
          "posting_about_hiring": boolean,
          "posting_about_qa_or_backlog": boolean,
          "stealth_mode": boolean,
          "too_big_for_icp": boolean,
          "too_small_for_icp": boolean
        },
        "icp_evaluation": {
          "cluster": "CTO | VP-Eng | Dir-Eng | QA-lead | CPO | VP-Product | PM-senior | Founder | COO | Investor | Other",
          "segment": "Founder-CTO-small | Scaleup-CTO | Enterprise-CTO | CPO-scaleup | Founder-PM | QA-lead-scaleup | not-icp | unclear",
          "company_size_estimate": "<50 | 50-100 | 100-1000 | 1000+ | unknown",
          "geography": "DACH | UK | NL | Nordics | US | Other | unknown",
          "score": {
            "title_fit": 0,
            "size_fit": 0,
            "geo": 0,
            "recency": 0,
            "warmth": 0,
            "total": 0
          },
          "tier": "A | B | C | D"
        },
        "outreach": {
          "suggested_angle": "builder-entry-buyer | operator-pm-entry | channel-white-label | advisor-ask-V0 | speaking-or-community | coordinate-before-touch | not-now",
          "opening_line": "one sentence tailored to this specific person — reference something concrete from their profile, not a template",
          "flags": ["string", "..."],
          "followup_notes": "2-3 sentence rationale for the tier and angle"
        },
        "confidence": 0.0,
        "extraction_notes": "anything the LLM could not reliably extract or had to infer"
      }
    as: assessment

  - type: control.stop
    message: '{{assessment}}'