Personal opencode CLI configuration directory (~/.config/opencode).
make install # installs snip (output compression tool)
npm install # installs @opencode-ai/plugin dependencyopencode.json — Core settings:
- Provider: vLLM (local model serving via OpenAI-compatible API)
- Model:
qwen3.6-27b(200k context, 40k output) - MCP Server: Context7 (documentation lookup, API key in
.context7-api-key) - Plugin:
@tarquinen/opencode-dcp(conversation state management) - Features: LSP enabled, formatter enabled
AGENTS.md — Agent behavior guidelines (personality, restrictions, tool usage conventions)
dcp.jsonc — Conversation persistence configuration
Custom agent skills in skills/:
| Skill | Purpose |
|---|---|
bootstrapping-plan |
Design new projects via iterative exploration |
create-command |
Create/modify opencode custom commands |
create-python-project |
Scaffold Python projects from cookiecutter template |
create-skill |
Generate new agent skills |
customize-opencode |
Edit opencode's own configuration files |
exploring-code |
Deep codebase exploration via subagent |
gh-cli |
All GitHub CLI operations (PRs, issues, actions) |
init-openspec |
Initialize OpenSpec in a project |
parsing-json |
JSON parsing/transforming with jq |
quality-checks |
Iterative test/fix loops until clean |
robs-design |
Rob's system design standard, used with openspec |
robs-theme |
Rob's color and font preferences |
searching-files |
File content search with rg (ripgrep) |
.llm_host— vLLM server base URL.openai_api_key— API key for vLLM provider.context7-api-key— Context7 MCP API key