Przejdź do treści
Biblioteka skilli
Produktywność osobista · Automatyzacja

linkedin-browser

/linkedin-browser

Czyta profil, wyniki wyszukiwania ofert i osoby na LinkedIn w przeglądarce zalogowanego użytkownika oraz wysyła zaproszenia, które użytkownik zatwierdził.

Repozytorium
talent-agent ↗
Licencja
Apache-2.0
Ostatnia zmiana repozytorium
07.11.2026

Ocena i bezpieczeństwo

100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
1
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
2
Przestarzałość

Brak sygnałów. Repozytorium zmienione 07.11.2026, bez odwołań do wycofanych modeli i API.

Licencja

Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.

Co robi i kiedy

Czyta profil, wyniki wyszukiwania ofert i osoby na LinkedIn w przeglądarce zalogowanego użytkownika oraz wysyła zaproszenia, które użytkownik zatwierdził.

Wyzwalacze z opisu

Meet the hiring team
Oryginalny opis (EN)

The only way this pack touches LinkedIn. Drives the user's own logged-in LinkedIn in a browser agent to read their profile, job search results with "Meet the hiring team", or people (alumni, hiring team, ex-colleagues), and to send connection requests the user approved by queue_id. Every mode accepts pasted LinkedIn page text instead. Use when another skill needs LinkedIn data, or the user says to send approved connection requests.

Jak działa

  1. Before any mode
  2. Budget (shared across all modes and skills)
  3. Modes
  4. Allowed clicks
  5. File formats
Sekcje z nagłówków SKILL.md, w oryginale · 60 linii, 885 słów
Podgląd SKILL.md
---
name: linkedin-browser
description: The only way this pack touches LinkedIn. Drives the user's own logged-in LinkedIn in a browser agent to read their profile, job search results with "Meet the hiring team", or people (alumni, hiring team, ex-colleagues), and to send connection requests the user approved by queue_id. Every mode accepts pasted LinkedIn page text instead. Use when another skill needs LinkedIn data, or the user says to send approved connection requests.
---
# linkedin-browser

LinkedIn's User Agreement prohibits automated access, and accounts get restricted for high volume (real example: ~190 profile views in a day plus 100+ invites over ~5 days → "suspicious activity" lock). This skill is opt-in, slow, and capped. The caps below are local ceilings this pack keeps, not limits LinkedIn guarantees are safe.

## Before any mode
1. **Paste path.** If `linkedin_browser` is not `true` in `profile/prefs.yml`, or no browser tool is available, ask the user to paste the page text (or a "Save to PDF" export) and parse that. Skip the browser steps below.
2. **Login.** The user logs in themselves in the agent's browser. Stop at any login, 2FA, captcha, or "verify it's you" screen and hand back.
3. **Budget.** Read `profile/linkedin-ledger.md` (create it from the format below if absent). Count today's (prefs `timezone`) and the last 7 days' actions. Done when you know the remaining budget for this mode; zero remaining → stop and say when it resets.

## Budget (shared across all modes and skills)
| Action | Per day | Rolling 7 days |
|---|---|---|
| search (people or jobs query) | 5 | — |
| job cards read | 25 | — |
| profile opens | 20 | 80 |
| connection requests (`attempting` counts) | prefs `connect_daily_limit` (20) | prefs `connect_weekly_limit` (60) |

Wait 20–60 s (randomized) between profile opens and 30–90 s between connection requests. Read result pages before opening profiles; stay on pages 1–2 of results.

## Modes
**profile** — open `prefs.linkedin_profile_url` (or the user's "Me → View profile"). Capture headline, experience (title, org, dates, bullets), education, certifications, skills. Return them to `setup-profile` as *unconfirmed* facts.

**jobs** — use LinkedIn Jobs search with the user's roles, location, and "Internship / Entry level" filters. For each card (≤25): title, company, location + workplace type, posted date, LinkedIn job id, and from the job page: the original employer/ATS apply link if shown, salary if shown, and the **"Meet the hiring team"** names + profile URLs. Return rows to `find-roles` with `source: linkedin`. "Easy Apply" postings: record them; applying is the human's job.

**people** — use LinkedIn's own search:
- Hiring team: the job page's "Meet the hiring team"; then People search `<company> "<team>" manager` or `<company> university recruiter`.
- Alumni: `linkedin.com/school/<school-slug>/people/` with keywords (`founder`, a company, a role) and the location filter.
- Ex-colleagues: People search filtered by past company (co-op employer, club, hackathon org) + location.
Capture per person: name, headline/title, current company, location, school or shared past org, mutual-connection count, canonical profile URL. Nothing else (no emails, phones). Return to `find-people`.

**connect** — sends approved connection requests. Requires `linkedin_connect: true`, `demo` not `true`, and a non-empty `linkedin_profile_url`.
1. Open "Me" and confirm the logged-in profile URL equals `prefs.linkedin_profile_url`. Mismatch → stop.
2. Read `profile/connect-queue.md`. If any row is `attempting` or `unknown`, stop and ask the user to check those profiles and tell you the real state.
3. Select rows in state `approved` whose `sender_url` equals the logged-in profile URL; skip any other row and report it. Approval exists only when the user named those `queue_id`s in chat in this conversation (e.g. "approve q3, q5"); record it as `approved_at` + the user's words. Approval covers sender, recipient, and note together: any change to `sender_url`, `recipient_url`, or `note` after approval puts the row back to `draft`.
4. For each row, within budget: open the recipient URL; if already connected or pending, set `skipped` with the reason; otherwise set `attempting` (write the file now), click Connect → Add a note, paste the **exact** approved note (or send without a note only if the approved note is `EMPTY`), click Send. Mark `sent` only after LinkedIn shows the invitation as pending; if you can't tell, mark `unknown` and stop the run. Log each action to the ledger.
5. Stop and hand back on: any warning or captcha, a weekly-invitation-limit notice, LinkedIn asking for the recipient's email, an out-of-personalized-notes notice (a note-less send needs new approval), or budget reached.
6. Report sent / skipped / failed / unknown rows and hand `sent` rows to `track-pipeline`.

## Allowed clicks
Navigation, search, filters, scrolling, opening profiles and job pages, and — in connect mode only — Connect, Add a note, Send for approved rows. Everything else is the user's: Message, InMail, Follow, Like, Endorse, Apply / Easy Apply, withdrawing invites, account settings.

## File formats
`profile/connect-queue.md` (one row per request; `draft-outreach` adds `draft` rows):
```
| queue_id | contact_id | recipient_url | sender_url | note | state | approved_at | approval | updated |
|---|---|---|---|---|---|---|---|---|
| q1 | li-jane-doe | https://www.linkedin.com/in/jane-doe/ | https://www.linkedin.com/in/me/ | Hi Jane, … (≤200 chars) | draft | | | 2026-11-07 |
```
States: `draft → approved → attempting → sent | failed | unknown | skipped`.

`profile/linkedin-ledger.md` (append-only, one line per action):
```
| at (local) | mode | action | target | result |
|---|---|---|---|---|
| 2026-11-07 10:02 | people | search | "University of Toronto" founder Toronto | 12 results |
| 2026-11-07 10:04 | people | profile_open | li-jane-doe | ok |
| 2026-11-07 10:20 | connect | invite | q1 | sent |
```
A missing or unreadable ledger blocks connect mode until the user confirms how many invites they sent in the last 7 days; record that as one line.

Wersje

W bazie nie ma innych skilli o nazwie linkedin-browser.

Migawka bazy z 13.11.2026 · wersja paczki 7fe3bbfd6b0d (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 100find-people
    Wskazuje rekruterów, menedżerów i absolwentów w firmie docelowej dla studenta, z LinkedIn i publicznych źródeł, i zapisuje kontakty.
  • 100connection-note
    Pisze krótkie, spersonalizowane notatki do zaproszeń na LinkedIn w ramach limitu znaków, dla osób szukających pracy.
  • 95linkedin-watch
    Czyta feed LinkedIn w przeglądarce użytkownika i wyciąga oferty pracy ukryte w postach, tworząc osobny raport bez zmian w rejestrze.
  • 90role-scan
    Przeszukuje LinkedIn w przeglądarce użytkownika pod kątem ofert pracy w jego mieście, dobiera je do faktów i szkicuje pierwszą wiadomość.
  • 95linkedin-job-search-skill
    Szuka ofert pracy na LinkedIn na podstawie CV i ogłoszenia wzorcowego, sortuje je według dopasowania i nie wysyła aplikacji.
  • 100job-linkedin
    Pisze pakiet profilu LinkedIn, ocenia go w skali 100, planuje posty i komentarze oraz sortuje wklejone wiadomości, a publikuje je użytkownik.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.