hj.skill

English | 中文

hj.skill is a small collection of reusable Codex skills distilled from real workflows.

The goal is simple: keep repeatable methods public, keep personal or company-specific context local, and let each user build their own private memory without copying long background prompts into every conversation.

Quick Start

1. Clone this repository.
2. Choose a skill folder from the list below.
3. Copy that folder into your Codex skills directory.
4. If the skill needs private local files (see the Setup column below), run its setup script from the workspace where you want private data stored. By default, it creates .hj-skill-local/<skill-name>/.
5. Fill the generated *.local.md files with your own context.
6. Invoke the skill in Codex with its $skill-name.

Example skills directory paths:

macOS / Linux: ~/.codex/skills/
Windows: C:\Users\<you>\.codex\skills\

Example copy target:

~/.codex/skills/hackathon-application-answerer
~/.codex/skills/recruiting-screening-assistant

You can also keep this repo as a development workspace and copy individual skill folders into Codex only when they are ready.

Included Skills

Skill	Use it for	Setup
hackathon-application-answerer	Drafting and improving hackathon, builder program, fellowship, and startup competition application answers.	powershell -ExecutionPolicy Bypass -File hackathon-application-answerer\scripts\reset-local-memory.ps1
recruiting-screening-assistant	Preparing recruiting phone-screen scripts, evaluating assignments and post-call notes, and maintaining a ranked candidate tracker.	powershell -ExecutionPolicy Bypass -File recruiting-screening-assistant\scripts\reset-local-context.ps1
candidate-fit-tracker	Evaluating candidates against real client role contexts with evidence-backed fit scoring, risks, next actions, and maintained screening notes.	powershell -ExecutionPolicy Bypass -File candidate-fit-tracker\scripts\reset-local-context.ps1
resume-batch-cleaner	Batch-cleaning resume folders by strict education and local background rules, deleting direct rejects and moving ambiguous files into check/.	powershell -ExecutionPolicy Bypass -File resume-batch-cleaner\scripts\reset-local-rules.ps1
candidate-comments-template	Turning a candidate resume and chat record into a concise, copy-ready comments template with background, motivation, compensation, level, and performance.	No setup required
candidate-client-handoff	Turning one candidate chat record and file into a client-ready Chinese handoff plus a validated 姓名-TTC.pdf attachment.	No setup required
daily-matching-feedback	Capturing daily recruiting match/recommendation feedback across companies, roles, and candidates, then producing an evening review document, a matching-mechanism feedback table, and calibration suggestions.	No setup required
extract-job-table-from-images	Extracting job listings from recruitment screenshots, photos, posters, chat images, or scanned pages into a filtered three-column 岗位 / 公司 / 地点 table.	No setup required
regional-ai-talent-research	Researching public AI/CS/robotics talent across universities in a selected region and delivering a source-backed, deduplicated, review-ready Excel workbook of faculty, researchers, postdocs, PhDs, alumni, and public contact details.	No setup required
ai-news-gossip-comic	Fetching and verifying recent AI news, rewriting it in a playful gossip style, and producing sourced news cards, storyboards, and multi-panel comic prompts.	No setup required
ttc-pdf-normalizer	Converting one candidate file into a validated PDF named 姓名-TTC.pdf while preserving the source.	No setup required

Read each skill's README or SKILL.md for its exact workflow and examples.

Common Usage Pattern

Most skills in this repo follow the same loop:

1. Initialize private local files from public templates.
2. Add your own background, examples, preferences, role context, or process details into .hj-skill-local/<skill-name>/*.local.md, or into another directory passed with -LocalRoot.
3. Ask Codex to use the skill, such as Use $hackathon-application-answerer to answer this form.
4. Review and edit the output before using it.
5. Save useful final versions and lessons back into local files.

Over time, the public skill stays clean while your local version becomes more personalized.

Skill Structure

skill-name/
  SKILL.md                  # required: the reusable skill instructions
  README.md                 # optional: English docs
  README.zh-CN.md           # optional: Chinese docs
  agents/
    openai.yaml             # Codex agent config
  references/               # optional: public templates and guidance
    public-template-or-guidance.md
  scripts/                  # optional: local-file setup/reset script
    setup-or-reset-script.ps1

Only SKILL.md is required; add references/ and scripts/ as the skill needs them. references/ contains public templates and reusable guidance. Private memory should live outside the installed skill folder, usually in .hj-skill-local/<skill-name>/ in your working directory.

Privacy Rules

Commit:

• SKILL.md
• README.md
• README.zh-CN.md
• agents/openai.yaml
• references/*.md when they contain only public templates or generic guidance
• reusable scripts that do not contain private data

Do not commit:

• local/
• .hj-skill-local/
• *.local.md
• *.private.md
• resumes, candidate notes, portfolios, screenshots, logs, generated files, or company-private process details

The root .gitignore is configured to ignore local private files.

Add A New Skill

When adding a new skill:

1. Put reusable instructions in SKILL.md.
2. Put public templates and generic guidance in references/.
3. Put real personal or company-specific context in local/*.local.md.
4. Add a small setup/reset script if local templates need to be created.
5. Validate the skill before publishing.

This keeps the repo useful to others while still supporting deeply personalized local workflows.