A personal account of why OpenCode in the terminal became the author's preferred AI coding tool on Arch Linux. Key reasons include no vendor lock-in (any model via a config change), the free `big-pickle` model for daily work, and two workflow conventions — AGENTS.md for per-project context and SKILL.md for reusable procedures stored at `.opencode/skills/<slug>/SKILL.md`. OpenRouter consolidates billing across providers. The setup covers GitHub repo management, writing, and light Python/Bash scripting.
Table of contents
Vendor Unlockedbig-pickle is Pretty GoodWorking with OpenCode: AGENTS.md & SKILL.mdSKILL.md for OpenCodeGet j3ffyang ’s stories in your inboxWhat I Usually Use OpenCode ForTwo Agents I Use DailyUnified Payment with OpenRouterConclusionQuestions this post answers
How does OpenCode's SKILL.md work and where do I put the file?
A skill in OpenCode lives at `.opencode/skills/<slug>/SKILL.md`. The file uses YAML frontmatter at the top — requiring at minimum a `name` field (lowercase-hyphenated, matching the folder) and a `description` that explains what the skill does and when to trigger it. OpenCode scans for `**/SKILL.md` up to the git worktree and loads the full body only when a task matches, turning a multi-step procedure into a single request. Developers building repeatable OpenCode workflows track patterns like this on daily.dev.
What is the big-pickle model in OpenCode and is it free?
big-pickle is a free model available through OpenCode, noted for being consistently fast compared to other free options like Nvidia Nemotron Ultra, which can be slow at times. It handles complex logic spread across large codebases — hundreds of Python, Bash, and other files — reliably enough for daily use without incurring API costs. Those evaluating free models for terminal-based AI coding find comparisons like this on daily.dev.