From get-me-a-job
Use this skill when the user wants to reach out to people in their network or find contacts at a target company. Triggers on phrases like "help me reach out to someone at [company]", "who should I talk to about this role", "write me a LinkedIn message", "help me cold outreach", "find someone at [company] to speak with", "how do I network for this job", "can you write a message to [person]", "help me do informational interviews", "who in my network works at [company]", or "help me get a referral". Also invoke when the user mentions a target company and seems to be in job-hunting mode — offer to help them find the right people to contact.
How this skill is triggered — by the user, by Claude, or both
Slash command
/get-me-a-job:network-outreachThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
This skill helps the user identify the right people to contact at target companies and craft outreach messages that actually get responses.
This skill helps the user identify the right people to contact at target companies and craft outreach messages that actually get responses.
First step every time: Read ~/.claude/get-me-a-job/references/network-context.md for context on the user's network and which communities they're part of. Also read ~/.claude/get-me-a-job/references/outreach-style-guide.md for their preferred outreach tone and style.
If these files are missing or empty, tell the user: "You don't have your network context or outreach style set up yet. Want to do a quick setup now?" Walk them through it (what communities are you in? what tone do you prefer?) and save the files. Or tell them to run /setup.
For any target company or role, help the user identify the right people. The best contacts are usually:
Warm path (always try first):
Read ~/.claude/get-me-a-job/references/network-context.md to understand which communities the user belongs to. These are their warmest paths.
Target contacts by role type:
Email finding: If Hunter.io or Apollo is connected, use it to find email addresses. Try the Email-Finder tool with the person's name and company domain, or Domain-Search to find contacts at a company. If no email tool is available, fall back to LinkedIn outreach and email pattern guessing ([email protected], [email protected], etc.).
CRITICAL: Do not write a single word of outreach before researching the company and the person.
Research the company:
Research the person:
Read ~/.claude/get-me-a-job/references/outreach-style-guide.md for the user's preferred approach. The default is the value-first product insight approach:
The value-first approach (default):
This works because it proves you can actually do the job. You're not saying "I'm curious about your search experience," you're saying "I used your search, here's what's broken, here's how I'd fix it." That's a work sample disguised as an email.
If research doesn't surface a product problem, fall back to the curiosity bridge:
Message structure (either approach):
Tone rules:
Subject line rule (for email): The subject line should create curiosity, not describe credentials.
If there's no response after 5-7 days, one follow-up is fine and often necessary:
After two messages with no response, move on. Don't burn the contact.
Once someone agrees to a call, help the user prepare:
Before showing any outreach draft to the user, run it through the humanizer skill. Outreach that sounds AI-generated gets ignored or, worse, damages the user's reputation. The humanizer catches em dashes, inflated language, sycophantic tone, and other tells. This is not optional — it's the difference between an email that gets a response and one that gets deleted.
After the humanizer pass and after the user has seen and approved the draft, save it as a Gmail draft first so they can also see it in their inbox before sending:
python ${CLAUDE_PLUGIN_ROOT}/lib/run.py gmail draft [email protected] "Subject" @/tmp/outreach-body.txt
Write the body text to a temp file (e.g. /tmp/outreach-body.txt) rather than passing a giant string on the command line — it's safer and survives shell quoting issues.
Send only when the user replies "send it" / "yes, send" (or equivalent unambiguous approval) in chat:
python ${CLAUDE_PLUGIN_ROOT}/lib/run.py gmail send [email protected] "Subject" @/tmp/outreach-body.txt
For follow-ups in an existing thread, use gmail reply MESSAGE_ID @body.txt instead — it reuses the original thread and Subject so the conversation stays threaded.
If the user wants to track outreach, optionally label the sent message:
python ${CLAUDE_PLUGIN_ROOT}/lib/run.py gmail label MESSAGE_ID "Outreach/Sent"
NEVER call gmail send or gmail reply without showing the user the full draft first and getting their explicit approval in the chat. Draft first, approve first, send second. No exceptions. A gmail draft (which only stages, doesn't deliver) is fine to run after the humanizer pass.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub kope-kope/haas-job-plugins --plugin get-me-a-job