agentleFS
eklavyagoyaleklavyagoyal/wingman/skills/apply/SKILL.md

apply

Skill0 starsChanged 2 months agoMITView on GitHub

Fill a German application form - Personio, softgarden, Greenhouse, Lever, Workday, Interamt

25 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use outreach, track and 8 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

  • 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

  • kstonekuanSkill

    Daily

    Documentation and capabilities reference for Daily

    Also in 3 projects that use this

  • julienamorganSkill

    profile

    Researches the user's market, competitors, and competitive positioning to build a complete ICP profile. Shows findings for user approval before proceeding. Triggers: /profile, "research my market", "who are my competitors". Reads: ./gtm/company.md, ./gtm/icp.md. Writes: updates ./gtm/icp.md with competitive positioning. Standalone: yes (asks for domain if company.md missing).

    Also in 3 projects that use this

  • genomewalkerSkill

    explore

    RLM-style recursive memory exploration - dynamically navigate the memory graph

    Also in 4 projects that use this

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 0 are distinctive, and 7 are very specific to this project.

  • 0

    Common

    Most of the 24 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 7

    Very specific

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

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

    If there is no evaluation.md, run /wingman:evaluate first.

    line 23
  2. Very specific:

    If application-data.md exists, load it.

    line 27
  3. Very specific:

    Before touching the form, tell the candidate the shape of it in two lines — shared/references/interaction.md asks for this on long flows, …

    line 57
  4. Very specific:

    Use read_page for the refs you interact with, and scout-form.js for what exists.

    line 83
  5. Very specific:

    Ask once, fill once. Cache new answers under a Custom Answers section in application-data.md so the next application reuses them.

    line 123
  6. Very specific:

    After approval, fill in one pass via scripts/fill-page.md.

    line 127
  7. Very specific:

    Write applied.md in the job folder: date, ATS, language, status (Submitted / Draft), what was attached, anything left manual.

    line 143

“Common” is measured: kept by at least 6 in 10 of the 24 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

ATS patterns: shared/references/ats.md · Interaction: shared/references/interaction.md · Work auth: shared/references/work-authorization.md

Fills application forms. Never submits. The user takes the last step, every time.

Step 0: Prerequisites

CV required. Load application-data.md if it exists (created in Step 2). Load profile.md for the German fields — Gehaltsvorstellung, Kündigungsfrist, Arbeitserlaubnis.

Step 1: Target job

  • URL → match against DATA_DIR/jobs/; load posting.md, evaluation.md, language.md, the CV and letter. No match → fetch the posting and create the folder.
  • last → most recently modified job folder; confirm which job.
  • current → use the active tab as-is; match its URL against saved folders for context.

If there is no evaluation.md, run /wingman:evaluate first. Filling a form for an unscored job means the user may be applying to something a 30-second check would have ruled out.

Step 2: Application data

If application-data.md exists, load it. Otherwise extract what you can from the CV and profile, present it for confirmation, and save:

# Application Data

## Persönliche Daten
- Vorname / Nachname / E-Mail / Telefon
- Straße und Hausnummer / PLZ / Wohnort
- Land: Deutschland

## Profile
- LinkedIn / Xing / GitHub / Portfolio

## Standardantworten
- Wie haben Sie von uns erfahren: [board actually used]
- Bereits bei uns beworben: Nein
- Gehaltsvorstellung: [X].000 EUR brutto p. a.
- Kündigungsfrist: [from profile]
- Frühestmöglicher Eintrittstermin: [computed]
- Arbeitserlaubnis erforderlich: [from work-authorization.md — NOT US sponsorship logic]
- Aufenthaltstitel: [permit type, if held]

## Freiwillige Angaben
- Schwerbehinderung: keine Angabe

No US EEO block. German forms do not carry voluntary race/ethnicity self-identification, and importing those defaults into a German application is wrong. Schwerbehinderung is voluntary and the candidate's disclosure to make — default to no answer unless they said otherwise.

Step 2b: Say what is coming

Before touching the form, tell the candidate the shape of it in two lines — shared/references/interaction.md asks for this on long flows, and this is the longest one:

Two stops from here. First the fill plan, everything I intend to put in every field, for you to correct. Then the form itself, filled and screenshotted, for you to submit.

I do not click Submit, tick the DSGVO box, or upload a file. Those are yours.

On a multi-page wizard, say how many pages. Someone who knows there are six pages will sit through six; someone who thought there was one gives up on page three.

Step 3: Navigate and scout

Browser setup per shared/references/web-extraction.md. Detect the ATS from the URL per shared/references/ats.md.

  • Personio (*.jobs.personio.de/.com) — native single page expected. Multiple upload fields for Lebenslauf, Anschreiben and Zeugnisse; build the whole attachment set. DSGVO consent is tenant-dependent: a checkbox on some, implicit-on-submit on others (SMIGHT) — either way the candidate's action.
  • softgarden (*.softgarden.io) — native. Quick-apply via LinkedIn/Xing also offered.
  • Lever (jobs.lever.co) — append /apply. Most automation-friendly.
  • Greenhouse — extract the iframe tokens, then navigate to the direct form URL (see the reference).
  • Workday / SAP SuccessFactors — account required. Tell the user to sign in themselves, then continue. Account creation is not something this tool does.
  • Interamt — account required, same handling. Expect the full Mappe plus Tarif and Schwerbehinderung fields.
  • Unknown — navigate, screenshot, identify. If unrecognizable, say so and ask.

Scout before filling — with the DOM, not the accessibility tree. Read ${CLAUDE_PLUGIN_ROOT}/tools/scout-form.js and evaluate it with javascript_tool. It returns the complete inventory: every field with its full label, which are required by label text, the select options, the upload fields, and separate consent and voluntary lists.

Do not scout with read_page alone. Measured against a Personio-shaped form, read_page(filter="interactive") returned 8 of 15 controls. The seven it dropped were every select, both file inputs, all three Schwerbehinderung radios, and the DSGVO consent checkbox — precisely the controls that must be surfaced to the user rather than answered. A scout that cannot see the consent box cannot report it, and "never tick a consent box" then holds only because nothing mentioned one was there.

Two further traps confirmed on the same form: read_page returns an empty tree until the pane has rendered, so an empty result means take a screenshot and re-read, never "this form has no fields"; and it named the date field from its TT.MM.JJJJ placeholder rather than its Verfügbar ab* (erforderlich) label, which loses the required marker.

Use read_page for the refs you interact with, and scout-form.js for what exists. Where they disagree, the DOM is right.

The German ATS notes in the reference are structural, not element-level — never assume a ref from the file.

Step 4: Generate what the form needs

  • CV — always. Exists in the folder? Use it. Otherwise run the tailor-cv workflow inline and show it before continuing.
  • Anschreiben / cover letter — only if the form has a field for it. Then the anschreiben workflow inline.
  • Zeugnisse — if the form accepts attachments, use the audited set. Unaudited Zeugnis → say so and offer /wingman:zeugnis before it goes out.

Step 5: Propose everything at once

Scan every field, then present one consolidated plan — never field-by-field questions:

Plan for the [Company] application ([ATS]):

From your data:
  Vorname: …   Nachname: …   E-Mail: …   Telefon: …
  PLZ / Wohnort: …

German fields:
  Gehaltsvorstellung: 75.000 EUR brutto p. a.
  Eintrittstermin: 01.07.2026  (today + your 3-month Kündigungsfrist)
  Arbeitserlaubnis erforderlich: Nein — Blaue Karte EU held
  Wie haben Sie von uns erfahren: StepStone

Needs your decision:
  DSGVO-Einwilligung — required to submit. This is your consent to their
  data processing; I will not tick it for you.
  Schwerbehinderung — voluntary. Currently: keine Angabe.

Manual upload (I cannot upload non-image files):
  Lebenslauf:    [path]
  Anschreiben:   [path]
  Zeugnisse:     [paths]

Approve, or tell me what to change.

Ask once, fill once. Cache new answers under a Custom Answers section in application-data.md so the next application reuses them.

Step 6: Fill

After approval, fill in one pass via scripts/fill-page.md.

Never tick the DSGVO consent checkbox on the user's behalf, even with general approval of the plan — it is a specific consent action and needs its own yes. Same for any terms or arbitration checkbox.

File uploads: the browser tools can only upload images, so PDF and DOCX uploads are the user's to do. Give the exact paths in Step 5 so they can upload while reviewing.

Multi-page forms (Workday, SuccessFactors, Interamt): fill, click Save and Continue, read validation errors, fix, advance. Never click a final submit.

Step 7: Review, then stop

Screenshot the completed form. Walk through what is filled and what is still missing. Then stop and ask for explicit confirmation.

Wingman does not click Submit, Send, Absenden, „Bewerbung senden", or „Bewerbung abschicken". Not with prior approval, not when the user says "just do it", not when the form looks obviously complete. Drafting is reversible; submitting is not. The user clicks.

Step 8: Log

Write applied.md in the job folder: date, ATS, language, status (Submitted / Draft), what was attached, anything left manual. Update tracker.md per shared/references/tracker.md with the status and a follow-up date using the German timing table.


Response Format

  1. Status — what was filled, what needs manual upload, whether submitted
  2. Files — paths written
  3. Next — follow-up date, or the next job

Permissions

{ "permissions": { "allow": [
  "Read(~/.wingman/**)", "Write(~/.wingman/**)", "Edit(~/.wingman/**)",
  "mcp__claude-in-chrome__*"
] } }

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.