Personal collection of Codex skills used to automate day-to-day software development workflows with an AI coding agent.
Each folder is a self-contained skill: it ships a SKILL.md (the skill's instructions) plus any scripts or config templates it needs. Skills are loaded by Codex from a skills directory and applied automatically when a task matches their description.
| Skill | Purpose |
|---|---|
| ticket-driver | End-to-end ticket handling: pull review feedback (GitLab MR / GitHub PR / Redmine journal) or implement a feature, run each project's validation gate, then report back (status, assignee, screenshot, PR links) and hand off to QA. Three modes per project: auto / propose (human approval via Telegram) / signal. |
| openimis | Standard git + PR workflow for OpenIMIS modules: feature branch from develop, single commit, push, PR with the official openIMIS template; Comoros and CSU fan-out workflows. |
| mi | Validation workflow for a Docker-based mobile-invoice stack (React + PHP): all tests and static analysis run inside the project's dev container. |
| flownote | Validation and conventions for an async FastAPI backend + React/Vite admin frontend (TDD, i18n, ruff/mypy, vitest). |
Skills are consumed by Codex, not installed as libraries. To use them locally, point your Codex config at this directory (or a copy of it) and describe the task — the matching skill is picked up automatically.
~/.codex/config.toml
skills = ["/path/to/codex-skills"]
my-skill/
├── SKILL.md # Skill instructions (required)
├── config.example.json # Config template — copy to config.json, never commit real secrets
└── scripts/ # Helper scripts used by the skill
- Security: this repo only ever contains templates and placeholders. Real credentials (Redmine API keys, Telegram bot tokens) are loaded at runtime from a local, git-ignored
config.jsonand must never be committed. - The project names, hosts and commands shown in the skills are personal defaults — replace them with your own.
- These skills are opinionated personal workflows; they are published as a demonstration of AI-assisted software development, not as a general-purpose product.
MIT