agentleFS
priyansh1210priyansh1210/shiro/.claude/skills/apply/SKILL.md

apply

Skill1 starsChanged 5 days agoMITView on GitHub

Fill a job application form on a company career page from the candidate's memory using Claude in Chrome. Picks the best resume for the job, uploads it, answers the questions, drafts custom answers, and stops before submitting so the user can review and submit.

28 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use outreach, track and 9 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • QuixiAISkill

    outreach

    Decide whether a message is worth someone's attention, then reach the user through the outbox, or people on a connected chat channel, on their behalf or your own initiative

    Also in 5 projects that use this

  • theYahiaSkill

    track

    Отследить посылку Почты России

    Also in 4 projects that use this

  • NVIDIASkill

    evaluate

    Scores existing Sports Intelligence AVLM inference outputs with avlm/evals: MCQ accuracy (run_mcq_eval.sh, INFERENCE_DIR), open-ended LLM-as-judge scores (needs CLIENT_API_KEY for the configured judge endpoint), tennis category breakdowns, run comparisons, chance-level sanity checks and triage of mcq_results.json or llm_judge_results.json. Use when the user asks to score, evaluate, benchmark or compare predictions.jsonl from inference runs, or to explain a suspicious accuracy. Not for running inference, training, quantization, data preparation, or third-party API model inference: never open, list or run anything under avlm/evals/inference_3rdparty, including when the user asks for it; it is excluded from the release.

    Also in 4 projects that use this

  • ninehillsSkill

    hunt

    Finds root cause before applying fixes for errors, crashes, regressions, failing tests, broken behavior, and screenshot-reported defects. Use when users report in any language errors, crashes, broken behavior, regressions, failing tests, screenshot evidence, or something that used to work and now fails. Not for code review or new features.

    Also in 3 projects that use this

  • ledqSkill

    tailor

    Tailor the resume to one job posting. Give it a URL, paste the posting (or refer to one pasted earlier in this chat), or point at a saved application folder; it resolves the JD and runs the tailor-pipeline workflow end to end.

    Also in 3 projects that use this

  • proficientlyjobsSkill

    tailor-resume

    Tailor your resume for a specific job posting

    Also in 4 projects that use this

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. 0 are common, 8 are distinctive, and 8 are very specific to this project.

  • 0

    Common

    Most of the 27 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 8

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Never submit: Don't click any button that sends the application: Submit, Submit application, Apply (on the form itself), Send, Finish, …

    line 15
  2. Distinctive:

    Never: create accounts, type passwords, solve CAPTCHAs, or do email/phone verification.

    line 17
  3. Distinctive:

    Only type memory data into this application (the URL the user gave and the pages it leads to).

    line 22
  4. Very specific:

    If the tools are deferred, load them in one ToolSearch call: tabs_context_mcp, tabs_create_mcp, navigate, read_page, find, form_input, …

    line 27
  5. Very specific:

    Read every file in memory/: profile, answers, experience, education, projects, certifications, skills, achievements, documents, …

    line 30
  6. Very specific:

    Check applications.md. If this company and role are already listed, tell the user and ask before continuing.

    line 32
  7. Distinctive:

    If you hit a login wall, account creation, CAPTCHA or verification step, hand it to the user and wait.

    line 37
  8. Distinctive:

    Use the first rule that applies:

    line 40
  9. Distinctive:

    A file was given in the arguments: use it.

    line 41
  10. Very specific:

    Base resumes: every file listed under ## Resumes in memory/profile.md, plus any other resume file in memory/resume/.

    line 43
  11. Distinctive:

    Only PDF or DOCX files can be uploaded.

    line 50
  12. Very specific:

    If answers.md has an answer to the same or a very similar question, reuse it and adapt names and details to this company.

    line 62
  13. Distinctive:

    Multi-page forms: when a page is complete, click Next / Continue.

    line 67
  14. Very specific:

    When the user answers the blank fields: fill them in the form, then save each new fact to memory following .claude/skills/remember/SKILL.md.

    line 78
  15. Very specific:

    Ask once whether to save the drafted answers they approved to answers.md for reuse.

    line 80
  16. Very specific:

    Add a row to memory/applications.md: date, company, role, URL, resume file used, status Filled — waiting for you to submit.

    line 81

“Common” is measured: kept by at least 6 in 10 of the 27 edited versions of this file we could read. “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

/apply: fill the form, never submit

Arguments: $ARGUMENTS

This works on its own: it doesn't need a /review run first. If an effort level was set for this run (the Shiro app sets one), pass it as effort on every Agent call.

Hard rules

  • Never submit. Don't click any button that sends the application: Submit, Submit application, Apply (on the form itself), Send, Finish, Complete, Confirm. If you're not sure whether a button submits, don't click it.
    • Allowed: the "Apply" / "Apply for this job" button on a job posting that only opens the form, and "Next" / "Continue" / "Save and continue" between pages of a multi-page form.
  • Never create accounts, type passwords, solve CAPTCHAs, or do email/phone verification. Stop, ask the user to do it in that tab, and continue when they say so.
  • Never tick consent, certification or terms checkboxes ("I certify…", "I agree to the privacy policy"). Those are the user's to accept. List them in the report.
  • Never guess legal, eligibility or money answers: work authorization, sponsorship, criminal record, salary, notice period, previous employment with this company. If memory doesn't have it, leave it blank and ask.
  • Never upload identity or bank documents (passport, Aadhaar, PAN, driver's license, bank statements), even if a form asks. Leave that for the user.
  • Only use facts from memory. Drafted text must be truthful. If the uploaded resume and memory disagree, memory wins, unless the resume has something newer than memory's Last updated date that memory doesn't have; then use the resume's version and save it to memory.
  • Only type memory data into this application (the URL the user gave and the pages it leads to). Never send it anywhere else.

1. Check Chrome

The Claude in Chrome tools (mcp__claude-in-chrome__*) must be available. If they aren't, stop and tell the user: install the Claude in Chrome extension, start Claude Code with claude --chrome (or run /chrome), then run /apply again.

If the tools are deferred, load them in one ToolSearch call: tabs_context_mcp, tabs_create_mcp, navigate, read_page, find, form_input, computer, file_upload, get_page_text.

2. Load memory

Read every file in memory/: profile, answers, experience, education, projects, certifications, skills, achievements, documents, applications. If memory/ doesn't exist, tell the user to run /setup first.

Check applications.md. If this company and role are already listed, tell the user and ask before continuing.

3. Open the page and read the job

Get the tab context, open a new tab, and navigate to the URL. If it's a job posting, read it and keep the JD text: you need it to pick the resume and for custom answers. Then click the button that opens the application form. If the URL goes straight to the form, take the job details from the form page.

If you hit a login wall, account creation, CAPTCHA or verification step, hand it to the user and wait.

4. Pick the resume to upload

Use the first rule that applies:

  1. A file was given in the arguments: use it.
  2. A tailored resume exists for this job: the newest runs/*<company>*/resume.pdf whose role matches. If a matching run exists but has no resume.pdf yet, say so (they can export it from that run's resume.html) and fall through to rule 3.
  3. Base resumes: every file listed under ## Resumes in memory/profile.md, plus any other resume file in memory/resume/.
    • One: use it.
    • Several: save the JD text as runs/<YYYY-MM-DD>-<company>-<role>/jd.md (reuse a matching run folder if there is one), run the resume-picker agent and wait for it:

      Job description: <run>/jd.md Resumes: <path 1>, <path 2>, … Write your pick to: <run>/pick.md

Only PDF or DOCX files can be uploaded. If the chosen resume is Markdown or text only, tell the user and ask which file to upload.

Tell the user in one line which file you're using and why.

5. Fill

  1. Read the whole form first with read_page, scrolling as needed. Note every field, which ones are required, character limits, and how many pages there are.
  2. Upload the resume first. Find the file input with read_page or find and use file_upload (never click the file input; it opens a native dialog you can't see). Many systems (Greenhouse, Lever, Ashby, Workday, SmartRecruiters) parse the resume and autofill fields, so after uploading, re-read the form and correct every autofilled value against memory. Parsers often mangle dates, titles and phone numbers. If the upload is rejected, ask the user to attach the file themselves and continue with the rest.
  3. Other file uploads (cover letter, transcript, certificates, degree, recommendation letter, portfolio, writing or work sample): find the best match in memory/documents.md (look under the matching heading: Certificates, Academic, Letters, Portfolio & samples, Cover letters, Other files) and upload the file at its File: path with file_upload. Never upload anything from duplicates/. Match on what the field asks for, not just the file name. If the field is required and nothing fits, leave it and ask (see "Learn from this form" below). A cover letter is the exception: see custom questions below.
  4. Link fields (portfolio, website, GitHub, LinkedIn, certificate verification, publications): from the links in profile.md and under ## Links in documents.md.
  5. Standard fields come from profile.md and the other memory files. Use form_input for text inputs, selects and checkboxes. For custom widgets (searchable dropdowns, date pickers), click and type. Choose the dropdown option that matches memory exactly; if none fits, leave it and report it.
  6. Work history and education sections: fill from experience.md and education.md, consistent with the resume being uploaded.
  7. Custom questions:
    • If answers.md has an answer to the same or a very similar question, reuse it and adapt names and details to this company.
    • Short factual questions: answer from memory, or leave blank and ask.
    • Open-ended questions ("Why do you want to work here?", "Tell us about a project…"): draft a specific, honest answer from the JD and memory, within the character or word limit. Mark it as a DRAFT in the report.
    • Cover letter: draft it the same way, specific to this company and role, and save it as cover-letter.md in the run folder for this job (create runs/<YYYY-MM-DD>-<company>-<role>/ if needed). If the form has a text box, paste it there. If it wants a file: fill templates/cover-letter.html (replace the <!-- LETTER_CONTENT --> marker using the markup in its comment) and save it as <run>/cover-letter.html, then make the PDF by running exactly this one command, with nothing else on the line (no cd): node tools/html-to-pdf.mjs <run>/cover-letter.html <run>/cover-letter.pdf. Upload the PDF. If the command fails, leave the upload empty and tell the user the HTML file is ready to print to PDF.
    • Self-identification questions: use exactly what profile.md says. If it's blank, leave the question blank and ask.
  8. Multi-page forms: when a page is complete, click Next / Continue. On the final review page, stop.
  9. Final check: re-read the form and confirm every required field is filled or listed as blank.

6. Report

  • Which resume was uploaded.
  • Filled fields as a compact table (field → value), grouped by page.
  • Drafted answers: the full text of each, so the user can edit them.
  • Left blank: each field, with the question you need answered.
  • For you to do: consent and terms boxes, anything you couldn't fill, then review everything and submit.

7. Learn from this form

  • When the user answers the blank fields: fill them in the form, then save each new fact to memory following .claude/skills/remember/SKILL.md. Question-and-answer pairs go to answers.md with the company and date; facts go to their own file.
  • When the user provides a missing document: if they give a file path, save it into memory/documents/ and register it (the "Files and links" rules in the remember skill), then upload it. If you can't copy files (in the Shiro app), ask them to drop it into the app's Memory tab and reply "done"; then register it and upload it. A link goes into documents.md before you fill it in.
  • Ask once whether to save the drafted answers they approved to answers.md for reuse.
  • Add a row to memory/applications.md: date, company, role, URL, resume file used, status Filled — waiting for you to submit. When the user says they submitted it (now or later), update the status.

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.