agentleFS
mukewlmukewl/jobkit/.claude/skills/apply-to-job/SKILL.md

apply-to-job

Skill1 starsChanged 3 months agoMITView on GitHub

End-to-end prep for one job application - read the posting, produce a tailored page-checked CV in the user's own formatting, find 5-10 named people worth contacting at that company, and draft LinkedIn messages for the user to send by hand. Use when given a job URL or job description and asked to apply, prepare an application, tailor a CV for a role, or find who to reach out to about a job.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. DhruvPrajapati4/apply-kit★ 0same name
    2. lecegues/hitapply-plugin★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository4 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 6 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read profile.md at the repo root before anything else.

    line 12
  2. Very specific:

    Stop and warn before doing any work: if the posting breaches anything in profile.md — a language above the stated level, the right to work …

    line 27
  3. Very specific:

    Follow .claude/commands/tailor-cv.md exactly.

    line 33
  4. Very specific:

    Use the Explorium (Vibe Prospecting) MCP.

    line 40
  5. Very specific:

    NEVER call export-to-csv: Not once, not "just this table", not if the credit balance looks healthy.

    line 58
  6. Distinctive:

    Never confirm any credit-consuming action on the user's behalf.

    line 62
  7. Distinctive:

    If nothing relevant surfaces after two rounds, say so plainly and fall back to the posting itself, the company's own team and about pages, …

    line 87
  8. Distinctive:

    Ask, do not sell: No metrics anywhere in outreach — no CAC figures, no revenue, no follower counts.

    line 138
  9. Distinctive:

    Never close with a challenge or an ultimatum.

    line 140
  10. Distinctive:

    Do not quote someone's own profile back at them.

    line 143
  11. Distinctive:

    If the workbook is locked (open in Excel), say so and ask the user to close it rather than silently writing somewhere else.

    line 146
  12. Very specific:

    Leave in applications/<Company>/: the CV PDF and its source.

    line 151

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Prepare one job application

Given a job posting, produce everything needed to apply by hand: a tailored CV, a shortlist of real people, and drafted messages. Nothing is ever sent automatically, and nothing that costs money is ever bought.

Read profile.md at the repo root before anything else. It carries work authorisation, languages, salary and the user's hard rules. If it is missing, stop and say so.

Output folder: applications/<Company>/

Step 1 — Read the posting

URL → fetch it. WebFetch first; when it 403s (Welcome to the Jungle and Ashby both do), open it in a browser tool and read document.body.innerText. If both fail, stop and ask for the pasted text. Never guess the role.

Extract: company, exact title, location, contract type, years required, responsibilities, must-have skills, language requirement, and any stated work-authorisation, citizenship or clearance requirement.

Stop and warn before doing any work if the posting breaches anything in profile.md — a language above the stated level, the right to work somewhere the user is not authorised, citizenship, or a clearance. Say it plainly and ask whether to continue.

Step 2 — Tailor the CV

Follow .claude/commands/tailor-cv.md exactly. It is the single source of truth for the CV rules: master is read-only, invent nothing, never exceed the profile's language level, keep what the profile says must survive, never touch the preamble, and verify the page count with cv/build.py — which exits non-zero on overflow.

Step 3 — Find 5-10 people worth contacting

Use the Explorium (Vibe Prospecting) MCP. Load its tools with ToolSearch first — search vibe prospecting fetch entities prospects and select fetch-entities and show-sample.

How this stays free

Explorium's search and preview are free; export costs credits. You never need the export.

show-sample returns five people with real names, titles and LinkedIn URLs already visible. Read them straight off that response and write them into Contacts.xlsx yourself. That is the entire technique. Want more than five? Run the fetch again with exclude_key: "prospects" and read the next five. Every round is free.

The paid export exists to hand you a CSV. You have an agent that can type, so you do not need one.

Absolute rules — money

  • NEVER call export-to-csv. Not once, not "just this table", not if the credit balance looks healthy. Search and preview are free; export is not.
  • NEVER call enrich-prospects. It returns masked values that only unmask on export, so it buys nothing here.
  • Never confirm any credit-consuming action on the user's behalf. If a tool asks to spend, stop and report.
  • cost_in_credits in a response is an export estimate, not a charge. Exploration already happened for free. Do not report it as money spent.

Procedure

  1. Find the company's business_id. fetch-entities with entity_type: "businesses", website_keywords set to the company name and its domain word, plus company_country_code. Confirm business_name and business_domain really are the employer before continuing — name collisions are common and cost you the whole run.
  2. Fetch prospects at that business_id. entity_type: "prospects", with:
    • business_id from step 1
    • job_department matching the role's function; add adjacent departments only when the JD calls for them
    • job_level — for a large company (1000+), use ["manager","senior manager","director"]. The VP and C-suite layer is too far above a manager-level application to be worth contacting. For a small company (under ~200), use ["founder","c-suite","director","manager"], because there the founder or department head genuinely is the hiring manager.
    • prospect_country_code for the country the role sits in
  3. Call show-sample with the returned table_name. Five rows come back with real names, titles and LinkedIn URLs.
  4. Want more? Re-run step 2 with exclude_key: "prospects" and call show-sample again. Repeat until you have 5-10 relevant people. Every round is free.
  5. If nothing relevant surfaces after two rounds, say so plainly and fall back to the posting itself, the company's own team and about pages, and web search.

Reading the results

  • Quote records_available as what you got. records_matching_filters is upstream headroom — never present it as data you hold.
  • Prospect fetches return no email or phone values, by design. Do not imply otherwise, and never guess an address and present it as found.
  • LinkedIn URLs come back obfuscated (linkedin.com/in/ACoAA...). They resolve in a logged-in browser, but give the name and title too so the user can just search.

Rank people by how close they sit to the actual role — the team the job describes working with, then that team's manager. Say which are relevant and which are noise. A brand director at a different division is not the right contact for a growth role, and saying so saves the user more time than one extra name.

Never log into LinkedIn, search it while authenticated, send connection requests, or send InMail. The user sends everything by hand.

Step 4 — Draft the outreach

Append rows to Contacts.xlsx at the repo root with openpyxl, preserving existing rows and formatting. Columns:

Company | Role applied | Priority | Name | Title | Why them | LinkedIn | Connection note (<300 char) | Message once connected | Status | Date sent | Follow-up due | Notes

Priority 1-3 = contact these, 4-5 = optional, 9 = surfaced but not relevant. Set Status to Not sent and leave Follow-up due alone — it is a formula. Give every person a row; write messages only for priority 1-3. Row height ~200 for rows carrying messages.

For each of the top 2-3 people write two things.

The connection note (under 300 characters, sent with the request): one line on what they applied for, one line on who they are. No metrics. No pitch.

The message once connected (under 140 words):

  1. Hi <Name>, thanks for connecting.
  2. The honest reason for writing — they applied, and would rather ask someone inside than work it out from the job ad.
  3. Two or three sentences on who they are, tied to something specific about that person's background. No numbers.
  4. One or two real questions — what the team is actually like day to day, plus something only that person could answer about their own path or the work.
  5. Would you have 15 minutes for a call in the next week or two? Happy to fit around whenever suits you.
  6. Thanks, / the user's first name

Ask, do not sell. No metrics anywhere in outreach — no CAC figures, no revenue, no follower counts. Those live on the CV, and reading as a pitch is what kills these messages. Never close with a challenge or an ultimatum. Where the user does not meet a requirement, name it plainly once and move on; that reads as honest, not weak.

Do not quote someone's own profile back at them. It reads as homework rather than interest. Noticing a job change or a career path is fine; reciting their About section is not.

If the workbook is locked (open in Excel), say so and ask the user to close it rather than silently writing somewhere else.

Step 5 — Deliver

Leave in applications/<Company>/: the CV PDF and its source. Contacts and messages go in Contacts.xlsx, not here.

Then report in chat, briefly:

  1. Output path and the confirmed page count
  2. What changed in the CV
  3. The people found, ranked, with one line on why each is or is not relevant
  4. The connection note and the message in full, so they can be copied straight from the conversation without opening the workbook
  5. Gaps — what the posting wants that the CV genuinely does not support

Finally, remind the user to log the row in Job_Tracker.xlsx, or offer to do it.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.