My personal OpenCode configuration, based on compound-engineering.

LSPs

Pyright

npm install -g pyright

Ruff

pip install ruff

Skills

All skills live in skills/ and are loaded on-demand via the skill tool. Each SKILL.md has a description field with trigger conditions — the LLM evaluates user intent against those descriptions to decide when to invoke.

Skill	Purpose
ce-strategy	Create or maintain STRATEGY.md
ce-ideate	Generate and critically evaluate grounded ideas
ce-brainstorm	Collaborative requirements exploration and dialogue
ce-plan	Create or deepen structured implementation plans
ce-work	Execute work — implement plans, ship features
ce-work-beta	Execute work with experimental Codex delegation
ce-debug	Systematically find root causes and fix bugs
ce-code-review	Structured code review with tiered persona agents
ce-resolve-pr-feedback	Resolve PR review feedback
ce-doc-review	Review requirements or plan documents
ce-compound	Document solved problems to compound team knowledge
ce-compound-refresh	Refresh stale learning docs in docs/solutions/
ce-commit	Create a git commit with a clear message
ce-commit-push-pr	Commit, push, and open a PR in one step
ce-frontend-design	Build web interfaces with genuine design quality
ce-optimize	Metric-driven iterative optimization loops
ce-simplify-code	Simplify recently changed code
ce-agent-native-architecture	Build agent-first applications
ce-agent-native-audit	Audit agent-native architecture with scored principles
ce-worktree	Create isolated git worktrees for parallel work
ce-product-pulse	Time-windowed pulse reports on product performance
ce-polish-beta	Iterate on UI improvements in a browser
ce-dhh-rails-style	Write Ruby/Rails in DHH/37signals style
ce-test-browser	Run browser tests on pages affected by a PR
ce-sessions	Search past coding agent session history
ce-session-inventory	Discover session files and extract metadata
ce-session-extract	Extract conversation skeleton from a session file
ce-clean-gone-branches	Clean up local branches whose remote is gone
lfg	Full autonomous engineering workflow

Model-restricted skills

These skills have disable-model-invocation: true in their frontmatter and an deny permission in opencode.json — the LLM cannot auto-invoke them:

• ce-polish-beta
• ce-agent-native-audit
• ce-work-beta
• lfg

Agents

Preconfigured subagents in agents/. These are dispatched by skills (e.g., ce-code-review, ce-doc-review) or manually via the task tool.

Category	Agents
Always-on reviewers	ce-correctness-reviewer, ce-testing-reviewer, ce-maintainability-reviewer, ce-project-standards-reviewer, ce-agent-native-reviewer, ce-learnings-researcher
Conditional reviewers	ce-adversarial-reviewer, ce-adversarial-document-reviewer, ce-security-reviewer, ce-security-sentinel, ce-security-lens-reviewer, ce-performance-reviewer, ce-performance-oracle, ce-api-contract-reviewer, ce-data-migrations-reviewer, ce-data-migration-expert, ce-data-integrity-guardian, ce-reliability-reviewer, ce-previous-comments-reviewer, ce-schema-drift-detector
Stack-specific reviewers	ce-dhh-rails-reviewer, ce-kieran-rails-reviewer, ce-kieran-python-reviewer, ce-kieran-typescript-reviewer, ce-julik-frontend-races-reviewer, ce-swift-ios-reviewer
Research agents	ce-repo-research-analyst, ce-web-researcher, ce-framework-docs-researcher, ce-best-practices-researcher, ce-session-historian, ce-issue-intelligence-analyst, ce-git-history-analyzer
Design agents	ce-design-iterator, ce-figma-design-sync, ce-design-implementation-reviewer, ce-design-lens-reviewer
Document reviewers	ce-coherence-reviewer, ce-feasibility-reviewer, ce-scope-guardian-reviewer, ce-product-lens-reviewer, ce-spec-flow-analyzer
Specialized	ce-architecture-strategist, ce-deployment-verification-agent, ce-pattern-recognition-specialist, ce-pr-comment-resolver, ce-code-simplicity-reviewer, ce-ankane-readme-writer

Sources

• https://github.com/compoundengineering/compound-engineering

Plugins

• https://github.com/Zeptiny/opencode-skill-injection-plugin

Third-party tools

• https://github.com/junhoyeo/tokscale

Changes from upstream compound-engineering

Removed skills

• ce-slack-research — Slack is not used
• ce-demo-reel — Not needed
• ce-report-bug — Not needed
• ce-setup — Not needed
• ce-release-notes — Not needed
• ce-proof — Not needed
• ce-gemini-imagegen — Not needed
• ce-test-xcode — Not needed

Removed agents

• ce-slack-researcher — Slack is not used

Skill description updates

All skill description fields now include trigger conditions ("Use when…") so the LLM can evaluate user intent without relying on a separate trigger table. Updated skills:

• ce-compound — added auto-invoke trigger phrases ("that worked", "it's fixed", "working now", "problem solved")
• lfg — added trigger phrases ("lfg", "full send", "ship it end-to-end")
• ce-work-beta — added Codex delegation trigger phrases ("use codex", "delegate to codex")
• ce-work — added execution trigger phrases ("do the work", "implement this", "execute the plan")
• ce-simplify-code — added simplification trigger phrases ("simplify this", "clean up this code", "reduce complexity")
• ce-agent-native-audit — added audit trigger phrases ("audit agent-native", "check agent parity")
• ce-polish-beta — added polish trigger phrases ("polish this", "iterate on this visually")
• ce-test-browser — added browser testing trigger phrases ("test this in the browser", "run browser tests")

Cross-reference cleanup

Removed all references to deleted skills from remaining files:

• ce-plan/SKILL.md and references/plan-handoff.md — removed Slack context dispatch and Proof HITL handoff
• ce-ideate/SKILL.md and references/post-ideation-workflow.md — removed Slack context section, Proof save/share/caller-aware return, Proof Failure Ladder
• ce-brainstorm/SKILL.md and references/handoff.md, references/universal-brainstorming.md — removed Slack context section, Proof HITL option and return-status handling
• ce-commit-push-pr/SKILL.md and references/pr-description-writing.md — removed ce-demo-reel capture flow and URL references
• ce-work/references/shipping-workflow.md — removed ce-demo-reel reference
• ce-work-beta/references/shipping-workflow.md — removed ce-demo-reel reference
• ce-frontend-design/SKILL.md — replaced /ce-setup install reference
• ce-test-browser/SKILL.md — replaced /ce-setup install references
• agents/ce-best-practices-researcher.md — removed ce-gemini-imagegen from tool routing

opencode.json

• Added deny permission for model-restricted skills (ce-polish-beta, ce-agent-native-audit, ce-work-beta, lfg)