OpenCode Skills
OpenCode Skills
Getting hands-on with AI-driven development means figuring out what actually works — and that looks different for every person, team, and codebase. There's no universal playbook here.
Over the past few months I've built Squirrel Notes, Fragile, this blog, and replaced my Adobe portfolio — all while evolving my approach to working with AI tools. I progressed from Claude Code to GitHub Copilot, then landed on OpenCode (while still using Copilot for some workflows).
The workflow I developed while building Fragile clicked for me, so I extracted it into a reusable set of skills you can find here.
The Skills
| Name | Description |
|---|---|
| architect | Drives technical design decisions, writes proposals before significant changes, and maintains the proposal index |
| developer | Writes production-quality TypeScript following TDD (red-green-refactor) and project conventions |
| reviewer | Reviews staged changes for security, correctness, performance, IaC safety, observability, and convention adherence; returns a PASS / PASS WITH COMMENTS / BLOCK verdict with Acceptance Criteria traceability |
| infosec | Read-only security and compliance audit (ISO27001-aligned by default). Audits encryption, access control, audit logging, secrets, IAM, network exposure, and supply chain. Returns APPROVED / REQUIRES CHANGES / APPROVED WITH EXCEPTION |
| decision-log | Captures and maintains architectural decisions (ADRs) in docs/decisions/ with a running index |
| dev-workflow | Full feature development cycle: proposal → implementation → review → infosec sign-off → decision logging → PR |
| project-bootstrap | Interactive bootstrap that asks a structured set of questions (app stack, IaC, observability, security/compliance, domain) and produces a complete CLAUDE.md and Project Context block |
| project-onboard | Interactive onboarding for an existing codebase — investigates the repo to fill in CLAUDE.md and the Project Context block, asking the user only what the code can't answer |
How It Fits Together
The core workflow I use on Fragile follows this chain: Architect → Developer → Reviewer → Decision Log. I've since added an infosec skill to close the loop on security sign-off before anything reaches a PR.
The project-bootstrap and project-onboard skills handle the setup side — configuring the right context for new projects or bringing existing codebases into the workflow. I initially built this blog and my photography site with minimal process, which made them a useful proving ground for the onboarding skill.
For personal projects at least, this approach with OpenCode is working well.
The Dev Workflow
The dev-workflow skill orchestrates the full feature development cycle — from proposal to merged PR — by sequencing the other skills in the right order and defining exactly when to loop back.