HireTy

An agentic workflow repository to help Tyson Lin get hired — faster and better.

The agent takes Tyson's professional profile, a job description, and produces a tailored CV (compiled to PDF via Typst) and a cover letter, all optimised through a pipeline of specialised sub-agents.

---

Repository structure

hire-ty/
├── data/
│   └── tyson/            # Tyson's professional profile, skills, and work history
│       ├── README.md     # How to maintain this folder
│       └── profile.md    # Master profile (fill this in first)
├── jobs/                 # One sub-folder per job application
│   └── README.md         # Folder naming convention and workflow
├── constraints/
│   └── content-guidelines.md  # Content rules all output must follow
├── subagents/            # Instructions for each specialised agent
│   ├── fact-checker.md
│   ├── think-like-tyson.md
│   ├── ats-analyzer.md
│   ├── keyword-gap-agent.md
│   ├── impact-quantifier.md
│   ├── tone-optimizer.md
│   ├── conflict-resolver.md
│   └── change-composer.md
├── Typst.md              # How to install Typst and compile CVs
└── README.md             # This file

---

Prerequisites

1. Fill in your profile

Open data/tyson/profile.md and complete every section with accurate, factual information. This is the only source of truth the agent uses — everything generated will be based on what you put here.

2. Install Typst (for compiling CVs to PDF)

See Typst.md for full installation instructions for macOS, Windows, and Linux.

Quick install:

# macOS
brew install typst

# Windows — download from https://github.com/typst/typst/releases

# Linux (Debian/Ubuntu)
sudo apt install typst

Verify:

typst --version

3. AI model / agent runner

This repository is designed to be used with an AI coding agent (e.g. GitHub Copilot Workspace, Claude, GPT-4, or any agent that can read files and follow instructions). No specific platform is required — the agent instructions are plain Markdown files that any capable model can follow.

4. (Optional) Set up PII Detection Pre-Commit Hook

To protect your personal information from being accidentally committed to Git, you can optionally set up an AI-powered pre-commit hook that scans for PII before each commit.

This is completely optional — your code will work fine without it. But if you're sharing this repo or want an extra safety layer, follow these steps:

# 1. Ensure Python 3 is installed
# macOS: brew install python3
# Linux: sudo apt install python3
# Windows: https://www.python.org/downloads/

# 2. Copy the example configuration
cp .env.example .env

# 3. Edit .env and add your API key (Anthropic or OpenAI)
nano .env

# 4. Install the required Python package
pip install anthropic  # or: pip install openai

That's it! The hook is already installed. On your next commit, it will scan for PII and block the commit if any is detected.

For full details, see .githooks/README.md.

---

How to apply for a job

Step 1 — Provide the job description

Give the agent a job description. Either:

• Paste the full text of the JD, or
• Provide the URL and ask the agent to fetch and summarise it.

Step 2 — The agent creates the job folder

The agent will create a folder under /jobs/ named:

<Company Name> - <Job Title> - <City>, <Country>

And populate it with:

File	Contents
job-description.md	Structured summary of the JD
cv.typ	Typst source for your tailored CV - this should be generated per job position
cover-letter.md	Tailored cover letter in email format

Step 3 — Review Initial Drafts

The agent has now created:

• cv.typ — Initial tailored CV (not yet optimized)
• cover-letter.md — Initial tailored cover letter (not yet optimized)

You can review these drafts and either submit as-is, or proceed to Step 4 to optimize them through the sub-agent pipeline.

Step 4 — (Optional) Run the sub-agent pipeline

When you're ready to optimize the drafts, ask the agent:

Run the sub-agent pipeline for [job folder name]

The agent runs these sub-agents in sequence. All agent output should be stored in /jobs/*/analysis/. The pipeline runs in four phases:

Phase 1: Critical Quality Gates (must pass before proceeding)

#	Sub-agent	What it does
1	Fact Checker	Verifies all claims against your profile (/data/tyson/profile.md). If it flags issues, the pipeline halts for corrections.
2	Think Like Tyson	Ensures the document reflects your authentic voice and personal guidelines. If misaligned, the pipeline halts for revision.

Phase 2: Functional Improvements (only if Phase 1 passes)

#	Sub-agent	What it does
3	ATS Analyzer	Checks keyword coverage and format compatibility
4	Keyword Gap Agent	Finds semantic mismatches between your profile and the JD
5	Impact Quantifier	Rewrites weak bullets into quantified achievements
6	Tone Optimizer	Aligns writing style with the role's seniority and company culture

Phase 3: Final Arbitration (only if Phases 1 & 2 complete)

#	Sub-agent	What it does
7	Conflict Resolver	Resolves contradictions between agents and recommends final changes (does not modify files directly)

Phase 4: Change Composition (only if Phase 3 complete)

#	Sub-agent	What it does
8	Change Composer	Drafts proposed CV and cover letter with all recommended changes applied; generates PDF for visual review

Step 5 — Review Proposed Changes

The Change Composer generates proposed files in the analysis/ folder:

File	Purpose
analysis/proposed-cv.typ	Full CV with recommended changes applied
analysis/proposed-cv.pdf	Compiled PDF of proposed CV for visual review
analysis/proposed-cover-letter.md	Full cover letter with recommended changes applied
analysis/change-summary.md	Clear summary of what changed and why

Review these proposed files and decide:

• Approve — Apply all changes to actual cv.typ and cover-letter.md
• Request modifications — Ask for specific iterations before applying
• Reject — Keep current versions unchanged

Step 6 — Compile and Submit

Once you approve the changes:

cd "jobs/<Company Name> - <Job Title> - <City>, <Country>"
# Copy approved changes from proposed files
cp analysis/proposed-cv.typ cv.typ
cp analysis/proposed-cover-letter.md cover-letter.md
# Recompile the final PDF
typst compile cv.typ cv.pdf

---

Interview preparation

Once you have an interview scheduled, ask the agent:

"Generate interview prep for [job folder name]"

The agent will create interview-prep.md inside the job folder with:

• Role-specific likely questions.
• Suggested answers based on your profile.
• Questions to ask the interviewer.
• Research notes about the company.

---

Research functionality

Research specific topics related to your job application to gain deeper context for interviews.

How to request research

Ask the agent:

Research [topic] for [job folder name]

For example:

• Research distributed systems architecture for [Company]
• Research financial services compliance for [Company]
• Research real-time data processing patterns for [Company]

What the agent creates

The agent conducts internet research and creates:

research/[topic]-research.md

Contains:

• Overview — High-level summary of the topic in simple, concise language
• Key Concepts — Main ideas relevant to the role and company
• Industry Context — How the topic fits into the broader industry
• Company Relevance — How this topic applies to the company's business or role
• Connection to Your Experience — Parallels with your profile and past projects
• Expected Knowledge & Behaviors — What the company expects someone in this role to understand
• Key Takeaways — Core insights you should be ready to discuss
• Glossary — Industry/domain-specific terms and acronyms
• External References — All sources and links used

Why research helps

Research reports:

• Provide deeper context for interviews and conversations
• Help you articulate connections between your experience and the role's requirements
• Identify gaps you should refresh before speaking with the company
• Give you concrete talking points that show genuine interest
• Build confidence when discussing technical or domain-specific topics

Research + Interview Prep

When you generate interview prep after completing research:

Generate interview prep for [job folder name]

The agent will automatically:

• Incorporate your research findings into suggested answers
• Reference research context when explaining expected knowledge
• Build talking points around your research insights
• Flag research topics as areas you're well-prepared to discuss

Research is separate from interview prep but works together with it.

---

Content rules

All generated content must comply with the rules in constraints/content-guidelines.md:

1. Factual accuracy — every claim must come from data/tyson/.
2. Concise, simple language — short sentences, active voice, no filler.
3. No AI-sounding language — no "leverage", "delve", "synergy", or similar; see the full banned list.

---

Updating your profile

Whenever your situation changes (new job, new skill, new certification), update data/tyson/profile.md. All future job applications will automatically reflect the change.

---

Troubleshooting

Problem	Solution
typst: command not found	See the Installation section in Typst.md
CV PDF looks wrong	Run typst watch cv.typ cv.pdf and check for errors in the terminal
Agent invents facts	Remind the agent to follow constraints/content-guidelines.md Rule 1
Missing keywords in CV	Re-run the ATS Analyzer sub-agent with the updated draft