Turn one current CV into a private, reusable application system for jobs, degrees, scholarships, fellowships, grants, conferences, workshops, research schools, hackathons, and competitions.
The kit creates the working data in your Google Drive, not in GitHub. It keeps a sourced ground-truth database, lane-specific tracker tabs, a separately tailored CV for every relevant opportunity, current materials, exact submitted Q&A archives, scheduled status monitoring, and a mandatory human review before any submission.
Set up my private application system using this repository:
https://github.com/Alcray/application-agent-kit
Install or activate all skills from the repository, then run the
application-agent-setup skill from beginning to end.
Requirements:
- Ask my name first.
- Ask which kinds of applications I will make: jobs/internships, degrees,
scholarships/fellowships, grants, conferences/workshops/research schools,
hackathons/competitions, or a custom type.
- Use my private Google Drive, Google Docs, and Google Sheets. If those tools
are not connected with write access, stop and help me connect them.
- Create the private folder structure, System Index, Ground Truth Database,
and only the tracker tabs needed for my selected application types.
- Ask me to upload my current CV into the created CV folder.
- Keep my general/master CV separate from Application CVs. For every
application that requests a CV or resume, create a new opportunity-specific
version without overwriting the master or a previously submitted version.
- Link the exact tailored CV from the tracker and preserve the uploaded version
with the submitted-application archive.
- Build a materials inventory for my selected application types, tell me what
useful foundational files are missing, and ask me to upload them into the
correct private folders. Do not request high-risk identity or financial
documents during general setup.
- Extract only facts supported by that CV, show me the inferred identity and
contact fields, explicitly confirm my legal given and family names, ask for
relevant missing facts, and record provenance for every value.
- Never put my CV, personal facts, Google Drive links, or credentials in the
cloned repository or any Git commit.
- Do not submit any application during setup.
- At the end, ask whether I want recurring application-status checks. If I do,
use the schedule-application-status-checks skill and ask for the frequency,
exact local time, timezone, and notification preference before scheduling.
When setup is complete, give me links to the root folder, System Index,
Ground Truth Database, tracker, and CV folder, plus a short list of anything
still missing.
That prompt is the recommended installation path because an agent can adapt it to the available skill installer and connector UI. The repository also follows the standard skill and plugin structure described in the official OpenAI skill documentation.
Application Agent — Your Name/
├── 00 System/
│ ├── System Index — Your Name
│ ├── Ground Truth Database — Your Name
│ └── Application Tracker — Your Name
├── 01 CVs & Resumes/
│ ├── General/
│ └── Application CVs/
├── 02 Supporting Materials/
├── 03 In Progress/
├── 04 Submitted Applications/
└── 05 Templates/
The tracker always has a dashboard and creates only the tabs you select. Each lane includes a Tailored CV link. The ground-truth sheet stores each fact with a verification state and source, plus a lane-specific materials inventory. The System Index lets the installed skills discover the right private files without modifying the public skill code.
application-agent-setup— asks the onboarding questions, builds or repairs Drive, reads the chosen CV, verifies gaps, and configures application lanes.tailor-application-cv— creates and visually verifies a separate truthful CV for one opportunity, records what changed, and never overwrites the master.apply-person— researches, fills, reviews, archives, and tracks a specific personal application. It always stops for a fresh, application-specific final authorization.check-application-status— checks email or portals for decisions and required actions, then reconciles evidence with the tracker without replying or accepting anything.schedule-application-status-checks— tests the checker, prevents duplicate schedules, and creates or manages a native recurring task at a confirmed time and timezone.
- An AI client that supports agent skills, or an agent capable of reading these
SKILL.mdfiles. - Google Drive, Docs, and Sheets access with permission to create private files.
- A controllable browser for application forms. The default configuration uses Chrome when available.
- Email access is optional for setup and application preparation, but required for automated status reconciliation.
Codex can install skills from another GitHub repository using $skill-installer. Ask it:
Use $skill-installer to install these paths from Alcray/application-agent-kit:
- skills/application-agent-setup
- skills/tailor-application-cv
- skills/apply-person
- skills/check-application-status
- skills/schedule-application-status-checks
Then use $application-agent-setup to set up my private application system.
The repo is also packaged as a plugin through .codex-plugin/plugin.json. Restart Codex if newly installed skills do not appear.
- Personal records stay in the user's private Drive.
- The repo contains schemas and instructions only.
- CV claims retain source provenance; ambiguous values stay
Candidateuntil confirmed. - Every tailored CV is a separate version tied to one opportunity; the general/master and submitted copies remain unchanged.
- Legal name, citizenship, residence, work authorization, and visa status are never guessed.
- Passwords, one-time codes, CAPTCHA responses, payment data, and government ID numbers are never archived.
apply-personprepares by default and does not submit until the user reviews the exact version and gives a fresh instruction naming that application.- Submission authorization never includes payment, enrollment, accepting an offer, signing terms, or another binding commitment.
After onboarding:
Use $apply-person to prepare my application for [opportunity URL].
To tailor a CV without filling the whole application:
Use $tailor-application-cv to tailor my CV for [opportunity URL].
After reviewing the completed browser form and exact snapshot:
Submit [exact application name].
To reconcile outcomes:
Use $check-application-status to check my applications and update the tracker.
To schedule a quiet daily monitor:
Use $schedule-application-status-checks to check my application statuses every
day at 09:00 in Europe/Yerevan. Notify me only when something changes or I need
to act.
Eligible ChatGPT Go accounts can use ordinary time-based scheduled tasks, including daily checks at an exact time. Current task limits and availability remain account-dependent; review the official scheduled-tasks documentation.
Go does not support event-triggered Gmail tasks, so this kit defaults to a normal time-based daily check. The scheduled run can inspect Gmail or another connected source only when that app is available and authorized for the account and surface.
The Plugin Directory is available across ChatGPT plans, but a particular plugin, skill, or app can still require a different plan or surface. Personal Skills in ChatGPT are generally documented for eligible managed workspaces. For a personal Go account, use this repository through Codex when Codex is included for the account, and verify that Google Drive, Sheets, email, and browser capabilities are actually available. See the current ChatGPT Go, plugin availability, and Codex plan documentation.
Run the repository checks:
python3 scripts/validate_repo.pyThe five skills are intentionally person-agnostic. Forks should keep private resource IDs and real applicant data out of Git.
MIT. See LICENSE.