LinkedIn Inbox

Most inboxes are mostly noise, and the cost is that the messages that matter wait a week. This skill separates them, then writes only what is worth writing.

Phase 1: Input

The user pastes the messages or screenshots. Never log into the account or read the inbox with a browser. Read the Voice section and the target roles in 01-Perfiles/ so recruiter messages can be judged against what the user is looking for.

Phase 2: Sort

Bucket	Signal	Action
RECRUITER	A role, a company, a salary band	Reply if the role is real and fits the target profile; one line if not
LEAD	Describes a problem the user solves, or asks about working together	Reply today, full answer
PEER	Same field, something to say	Reply this week, keep it human
ASK	Advice, time, an intro, a favour	Reply if it is cheap and specific; decline cleanly if not
SPAM	Agency pitch, lead-gen sequence, crypto, "quick question" with no question	Archive, no reply

Print the counts first: "3 recruiters, 1 lead, 41 spam" is most of the value.

Phase 3: Spot sequences

Automated outreach has a shape: an invite with no specifics, a message minutes after the accept, "quick question", "I noticed you're in {industry}", a calendar link in message one, a bump exactly a few days later. Mark it SPAM and say which tell gave it away. The user does not owe a reply to a script.

Phase 4: Replies

• RECRUITER, role worth it: ask what the message left out (salary band, level, remote or office, the stack) and say what makes the user a fit in one line. Log it in 05-Contactos/ and, when it becomes an application, hand off to /job-aplicar.
• RECRUITER, not a fit: one line: not looking for this, happy to refer, and mean the refer.
• LEAD: answer the actual question in full; if it fits, the offer is one sentence at the end.
• ASK: under ten minutes and specific: do it. "Can I pick your brain": decline in one warm sentence and give the one answer you would have given on the call.
• Declines are short, warm and final. No "let's revisit in Q3" if there is no Q3.
• Never invent facts about the user's work. Stack, numbers, timelines and how something was built come from 01-Perfiles/Perfil Maestro.md or from the user. When a reply needs one that is missing, write {{...}} in its place and ask; a made-up detail in public is worse than a shorter reply.

Run every draft through /content-humanizer (.claude/skills/content-humanizer/scripts/humanize.py, then detect.py from the same folder).

Phase 5: Output

INBOX  ·  52 items  ·  2 RECRUITER, 1 LEAD, 4 PEER, 2 ASK, 43 SPAM

RECRUITER
@Ana López - Senior Backend at Acme
> Thanks, Ana. Sounds close to what I'm looking for. What's the salary band, and is it
> remote? I've spent the last two years on Go services at a similar scale.

SPAM  (43) - archive. 38 are the same sequence: no-specifics invite, "quick question"
within 4 minutes of the accept, calendar link in message one.

Protocol

• The user sends every reply. The User Agreement (section 8.2) forbids bots and automated methods that send messages or add contacts.
• Ask before writing to 05-Contactos/. Never store salary details or contact data in files that go to a remote without explicit confirmation.
• Credit: adapted from Jake Schincariol's linkedin-agent-skill (MIT).

Next steps

• /job-aplicar when a recruiter role turns into an application.
• /cold-outreach for the follow-up if a recruiter goes quiet.