mzupan/tech-resume

A shared Agent Skill for Claude Code, Codex, IBM Bob, GitHub Copilot, Gemini CLI, and OpenCode that writes tailored resumes and cover letters for tech jobs. It interviews you once to build a career profile, asks a few targeted questions for each job description, and produces DOCX and PDF files that:

• parse cleanly in ATS and AI screeners (Greenhouse, Lever, Ashby, Workday): single column, contact info in the body, no tables or text boxes, embedded fonts, no ligatures
• cover the job description's must-have terms with bullets that prove them, not just a skills list
• don't read as AI-written: a lint pass flags stock AI resume words, dashes, uniform bullets, and -ing riders before anything is built

It never invents numbers, titles, or tools. If a bullet needs a fact you haven't given it, it asks.

Requires an agent that can read files and run shell commands, Python 3.9+, and git for cloning.

All supported agents use the same SKILL.md, scripts, references, and templates. Separate copies of the instructions for each agent are unnecessary; only the discovery location and invocation differ.

One installation under ~/.agents/skills/tech-resume serves Codex, Copilot, Gemini CLI, and OpenCode. Bob and Claude Code use their own discovery folders.

1. Choose a folder from the table and clone the skill there. For Codex (and the other agents that support the shared folder):

   SKILL_DIR="$HOME/.agents/skills/tech-resume"
   mkdir -p "$(dirname "$SKILL_DIR")"
   git clone https://github.com/mzupan/tech-resume.git "$SKILL_DIR"

   For Bob, use SKILL_DIR="$HOME/.bob/skills/tech-resume"; for Claude Code, use SKILL_DIR="$HOME/.claude/skills/tech-resume".

   To install the version in an existing checkout instead of cloning, run this from the checkout after setting SKILL_DIR to a new installation folder outside it:

   mkdir -p "$SKILL_DIR"
   cp SKILL.md README.md requirements.txt LICENSE "$SKILL_DIR/"
   cp -R scripts references templates "$SKILL_DIR/"

   For project installation, put the whole skill folder in the chosen project's discovery folder. A SKILL.md at the root of an ordinary checkout alone is not a discovery installation.

2. Install the Python packages the build and verify scripts use, in the same Python that runs the scripts:

   python3 -m pip install -r "$SKILL_DIR/requirements.txt"

3. Install poppler for pdftotext and pdftoppm (recommended; text extraction falls back to pypdf without it, but PNG previews require poppler or another PDF renderer):

   brew install poppler          # macOS
   sudo apt install poppler-utils  # Debian/Ubuntu

4. Optional: install the two writing skills the human-voice pass uses beside tech-resume in the same discovery folder. Without them, it uses its own writing rules and linter.

   git clone https://github.com/blader/humanizer.git "$(dirname "$SKILL_DIR")/humanizer"
   git clone https://github.com/theclaymethod/unslop.git "$(dirname "$SKILL_DIR")/unslop"

   • humanizer rewrites AI writing patterns based on Wikipedia's "Signs of AI writing".
   • unslop adds a phrase scanner that scripts/lint.py merges into its report.

5. Start a new agent session or reload its skills. In Bob, check Settings > Skills; in Gemini CLI, use /skills list.

PDFs use Arial when it's installed (macOS has it; on Linux install ttf-mscorefonts-installer) and fall back to Helvetica otherwise.

Run your agent from the folder where you want your career data to live (for example ~/career). The skill creates profile/ and applications/ there.

In Claude Code, invoke /tech-resume; in Codex, invoke $tech-resume. For Bob, Copilot, Gemini CLI, and OpenCode, ask in plain language, for example: Use the tech-resume skill to init old-resume.pdf. The operations below are the same in every agent; /tech-resume is shown as Claude Code syntax.

Each application gets its own folder:

applications/2026-10-11-company-role/
  jd.md  keywords.txt  gaps.md  resume.md  cover-letter.md  report.md
  First-Last-Resume-Company.docx/.pdf
  First-Last-Cover-Letter-Company.docx/.pdf

report.md lists must-have keyword coverage, what was left out on purpose, and the facts that would most improve that resume.

Your profile and applications hold personal data. Keep that folder out of public repos.

Installation makes the workflow available for automatic discovery and explicit skill invocation across sessions. It is useful if you tailor resumes repeatedly or want the same process across agents.

For a one-off task, an agent with file and shell access can use an ordinary checkout without installing it: ask it to read the checkout's SKILL.md and follow it, using the checkout as SKILL_DIR and a separate career workspace for your data. The scripts also work directly without an agent. Skill installation supplies reusable instructions; it does not add model capabilities or replace the Python dependencies.

The skill drives these, but they also work on their own:

references/research.md lists the sources behind the formatting and writing rules.

MIT