opencode-agent-orchestration-kit

OpenCode Agent Orchestration Kit

A reproducible product-development workflow for OpenCode.

Turn OpenCode into a structured team of specialized agents that can research, design, specify, implement, review, and validate software changes — without building every agent, command, skill, tool, and workflow convention from scratch.

Built for developers and small teams who want more structure than a single general-purpose coding agent, while keeping the setup local, inspectable, version-controlled, and adaptable.

[!NOTE] This project is a configuration and workflow kit for OpenCode. It is not a hosted platform, a fully autonomous software factory, or a replacement for engineering judgment.

Why this exists

A basic coding-agent setup often starts simple:

• One general-purpose agent.
• A few custom prompts.
• Some copied skills.
• Ad hoc instructions.
• Manual review and validation.

Over time, that setup becomes harder to maintain:

• Agents overlap or contradict each other.
• Planning, research, implementation, and review blur together.
• Workflows become prompt-dependent and difficult to reproduce.
• Configuration changes become risky.
• Documentation and validation drift away from the actual setup.

This kit provides a documented OpenCode harness with:

• Specialized agents with explicit responsibilities.
• Predictable handoffs between research, design, specification, implementation, and review.
• Slash commands for common product-development workflows.
• Local process skills and reusable engineering checklists.
• Safe installation and uninstall scripts.
• Local validation for configuration and harness contracts.
• Optional integrations for Open Design, Superpowers, Impeccable, and observability.

The goal is not to force every request through a heavy process. Small, clear, low-risk changes can go directly to implementation. Larger or ambiguous work can follow a structured flow with research, planning, review, and evidence.

What you get

Capability	What it gives you
Role-based agents	Dedicated agents for routing, research, design, specification, implementation, review, and optional harness evolution
Product-development workflows	Structured flows for features, plans, scoped research, bounded engineering loops, MVP specs, testing, simplification, and review
Bounded routing	A default lead agent that routes simple requests directly and escalates only when needed
Local skills	Practical checklists for testing, debugging, security, performance, documentation, APIs, code review, and more
Safe installation	Backup-aware install and uninstall scripts, plus a project-local test mode
Validation	Mechanical checks for JSON configuration, agent contracts, command contracts, and harness consistency
Adversarial safety	Portable fixtures for injection, unsafe paths, wrapper/network permissions, approval integrity, supply-chain pins, and corrupt events
Optional design workflow	Open Design integration for editable UI workspaces and design-oriented flows
Optional token visibility	TUI plugin for lead and subagent token usage when OpenCode exposes session trees
Versioned documentation	Agent, command, evidence, and validation contracts stored alongside the configuration

Built around

This kit builds on and integrates with the following projects:

• OpenCode — the coding-agent runtime and configuration surface.
• Open Design — optional local-first design workspace integration.
• Superpowers — optional agentic skills and development methodology.
• Impeccable — optional design guidance for AI-generated frontend work.

Each integration is optional. The core workflow can be used with OpenCode alone.

Who this is for

This kit is a good fit if you:

• Use OpenCode and want a repeatable development workflow instead of isolated prompts.
• Want specialized agent roles without building the whole system from zero.
• Want to separate research, planning, implementation, review, and validation.
• Prefer local configuration that can be inspected, customized, committed, and evolved.
• Work alone or in a small technical team.
• Want a practical starting point rather than a black-box orchestration platform.

Who this is not for

This kit may not be the right fit if you:

• Only need one small custom prompt or agent.
• Do not use OpenCode.
• Want a hosted multi-agent platform with queues, dashboards, billing, and team administration.
• Need autonomous parallel execution across many repositories or worktrees.
• Expect every request to be fully automated without human oversight.

How it works

Free-form requests start with the lead agent.

lead acts as a bounded router:

• Simple and low-risk implementation work goes directly to developer.
• Technical uncertainty goes to researcher.
• Visual or interaction work goes to designer.
• Planning gaps go to specifier.
• Larger feature work follows a structured orchestration flow.

Simple change
lead
  └── developer
        └── validation

Feature work
lead
  ├── designer            when visual or interaction design is needed
  ├── researcher          when technical uncertainty must be resolved
  ├── specifier           creates tasks, acceptance criteria, and validation plan
  ├── developer           implements approved work
  └── reviewer            checks the final diff against the agreed scope

Plan-only work
lead
  └── researcher
        └── specifier
              └── reviewer

Bounded engineering loop
lead
  └── human approval gate
        └── developer
              └── reviewer
                    └── developer state sync

Scoped research and specification
scoper
  └── researcher
        └── scoper synthesis
              └── specifier

Optional harness evolution
evaluator
  └── debugger
        └── evolver
              └── lead approval
                    └── developer
                          └── evaluator
                                └── debugger
                                      └── reviewer

The normal workflow is intentionally not a fixed ceremony for every task.

Use the smallest useful flow:

• Direct implementation for a small change.
• /plan when you want an implementation-ready plan but no code changes.
• /scope when you need research and an MVP-oriented specification.
• /feature when the work deserves full orchestration.
• /loop when you want an approved, resumable implementation-review cycle with crash-recoverable structured state and a three-iteration cap.
• /autonomous for one explicitly requested local task with a task-specific one-to-six-iteration budget, a hard ceiling of six, durable state, and reviewer-only closure. Final approval stops the cycle immediately.

Included agents

Agent	Responsibility
lead	Default router, feature orchestrator, and phase-barrier owner
scoper	Lightweight research-to-spec orchestration
designer	Product and interaction design using project context, optional Impeccable, and Open Design
researcher	Code, documentation, API, alternative, and risk investigation
specifier	Specs, tasks, acceptance criteria, scope boundaries, and validation plans
developer	Implementation, direct-mode changes, and focused validation
reviewer	Final, read-only review against scope, real diff, original evidence, and contextual architecture
evaluator	Optional benchmark and smoke-test evidence collection
debugger	Optional root-cause analysis from failures and traces
evolver	Optional evidence-driven harness improvement proposals

The evaluator, debugger, and evolver agents are optional harness-evolution sidecars. They are not part of the normal feature-development path.

Included commands

Command	Use it when you want to...
/feature	Run the full feature workflow
/plan	Research and create an implementation-ready plan without implementing
/scope	Research a topic and produce a scoped MVP specification
/loop	Design and run a bounded, resumable engineering loop
/mvp-spec	Create a strict MVP spec with small tasks and explicit out-of-scope items
/design	Create or evolve a design through project context and Open Design
/research	Run a direct research task
/spec	Run a direct specification task
/implement	Implement approved work directly
/test	Reproduce a bug or run focused validation
/code-simplify	Simplify code without changing behavior
/review	Run the final read-only review of the current diff
/evolve	Run the optional harness-evolution workflow
/init	Initialize project-oriented context for the workflow

Examples:

/feature Create onboarding with plan selection and welcome screen

/plan Add a dry-run flag to the harness check without implementing it yet

/scope Research Stripe Checkout integration and generate an MVP spec

/loop Add request retries with focused tests and independent review

/loop resume request-retries

/mvp-spec Email notifications when an agent finishes a task

/design Read PRODUCT.md and DESIGN.md, create an editable Open Design project, and generate a first version

/test Reproduce the checkout regression with a focused test

/code-simplify Simplify the parser branch without changing behavior

/review

/loop stores its approved human-readable contract in Markdown and uses a schema-versioned JSON snapshot, append-only JSONL history, and exclusive lock for execution state. Contract hashing rejects stale approvals, action IDs make retries idempotent, and explicit repair recovers interrupted writes without turning the loop into unattended automation.

A small request can also be written directly:

Change the Settings heading to Account settings and run the smallest relevant validation.

One review policy across stacks

All review entry points use the same language- and framework-neutral policy. The reviewer checks correctness, security, contracts, architecture, maintainability, tests, and evidence—not only whether tests pass.

Generic backend and frontend profiles adapt those checks to the changed surface. Strict Clean Architecture, DDD, hexagonal, CQRS, or layered profiles activate only when the task or repository explicitly declares that architecture; folder names alone do not activate them.

Only problems introduced or materially worsened by the current task can block it. Relevant pre-existing debt is reported separately as a non-blocking observation. /review-preflight remains deterministic, and focused /review-orchestrated modes remain partial; only the general /review agent emits a final verdict.

For this kind of request, lead should choose the direct path and delegate to developer without invoking the full feature workflow.

Quick start: test it safely first

The safest way to try the kit is to load it from the repository directory without changing your global OpenCode configuration.

Requirements

• Git
• Bash
• OpenCode installed and available on your PATH
• Node.js ^22.9.0 || ^24.0.0 and npm
• An OpenCode-compatible model provider configured through opencode auth login

1. Clone the repository

git clone https://github.com/jcarlosrodicio/opencode-agent-orchestration-kit.git
cd opencode-agent-orchestration-kit

2. Configure models

cp env.example .env
source .env

env.example provides a shared default model plus optional per-role overrides.

export OPENCODE_MODEL="openai/gpt-5.5"

export OPENCODE_SMALL_MODEL="$OPENCODE_MODEL"
export OPENCODE_LEAD_MODEL="$OPENCODE_MODEL"
export OPENCODE_SCOPER_MODEL="$OPENCODE_MODEL"
export OPENCODE_DESIGNER_MODEL="$OPENCODE_MODEL"
export OPENCODE_RESEARCHER_MODEL="$OPENCODE_MODEL"
export OPENCODE_SPECIFIER_MODEL="$OPENCODE_MODEL"
export OPENCODE_DEVELOPER_MODEL="$OPENCODE_MODEL"
export OPENCODE_REVIEWER_MODEL="$OPENCODE_MODEL"
export OPENCODE_EVALUATOR_MODEL="$OPENCODE_MODEL"
export OPENCODE_DEBUGGER_MODEL="$OPENCODE_MODEL"
export OPENCODE_EVOLVER_MODEL="$OPENCODE_MODEL"

You can use one model for every role or assign different models depending on cost, speed, and task complexity.

Interactive OpenCode model switching

The repository also includes oc-switch, a standalone TUI for changing the OpenCode default model, small/title model, and per-agent overrides. It builds its provider list from OpenCode itself and reads the local model catalog from opencode models --pure, so it only shows providers configured in your OpenCode installation. Startup is cache-first; press r when you want an explicit remote catalog refresh through --refresh. It does not share a catalog with Pi.

From this checkout, run:

npm run oc-switch

To install the command globally with npm from the cloned checkout:

npm install --global .
oc-switch

After a tagged release is published, you can install the same CLI directly from GitHub without keeping a checkout:

npm install --global "git+https://github.com/jcarlosrodicio/opencode-agent-orchestration-kit.git#v1.0.39"
oc-switch

The package exposes oc-switch through its npm bin entry; no separate wrapper or shell alias is required. Keep ~/.npm-global/bin (or your npm global bin directory) on PATH if npm is configured with a user prefix.

oc-switch quick guide

• ↑/↓ moves through Default, Small/title, and the agent rows.
• Enter or / opens the model picker for the focused row.
• Type to filter the current OpenCode catalog; q, r, and s are search characters while the picker is open. Backspace deletes one character.
• Enter applies the highlighted model. Space marks several agent rows so one selection can be applied to all of them.
• s saves. q saves and exits. Esc returns to the assignment list, and r explicitly refreshes OpenCode's catalog.

The selected values are persisted in the OpenCode model-switcher state and in the managed model block of .zshrc or .bashrc.

Pi has its own independent pi-switch; its catalog comes only from Pi's pi --list-models command.

3. Install local plugin dependencies

(cd opencode && npm ci --ignore-scripts)

4. Load the kit without global installation

export OPENCODE_CONFIG_DIR="$PWD/opencode"
opencode auth login
opencode

OpenCode now loads the configuration from this repository instead of your default global config.

Try one of these commands:

/scope Research whether this repository should use Stripe Checkout or Payment Element and produce an MVP spec

/plan Add a dry-run flag to the harness check without implementing it yet

/feature Add a small settings page with a saved theme preference

Install globally

Once you are happy with the workflow, preview and install it into your OpenCode configuration directory:

./install.sh --dry-run
./install.sh

The default target is:

${OPENCODE_CONFIG_DIR:-$HOME/.config/opencode}

Use a custom target when needed:

./install.sh --target "$HOME/.config/opencode"

An explicit --target wins over a non-empty OPENCODE_CONFIG_DIR, which wins over $HOME/.config/opencode. Argument values are literal; ~ is not expanded.

The installer inventories the payload, computes a complete plan, and tracks ownership in TARGET/.oak/manifest.json. A dry run performs no writes. A real operation recomputes the plan under an exclusive lock and commits through a durable journal with one rollback generation.

The root package.json is the canonical kit-version source. Display the same identity through the manager or any lifecycle wrapper:

node scripts/version.mjs
./install.sh --version
./upgrade.sh --version
./doctor.sh --version
./uninstall.sh --version
./rollback.sh --version

Safe defaults

Initial install preserves differing AGENTS.md, opencode.json, tui.json, package.json, and package-lock.json as user-owned files. Exact existing matches may be adopted without rewriting. Any other collision blocks the whole operation. A legacy installation without a manifest follows these same initial-install rules.

Use --force only on initial install to authorize replacement of colliding regular files after their bytes and modes are durably backed up:

./install.sh --force

It never authorizes directory or symlink replacement, unsafe paths, state repair, upgrade conflicts, or deletion of user-modified files.

After global installation, install the OpenCode config dependencies:

(cd "${OPENCODE_CONFIG_DIR:-$HOME/.config/opencode}" && npm ci --ignore-scripts)

Use that frozen command only when the installed package.json and package-lock.json still match the shipped manifests. If installation preserved differing manifests, review and merge the required dependencies first, then choose the install command for that user-owned dependency graph.

Upgrade, diagnose, uninstall, and rollback

The package also exposes the dependency-free oak CLI as a thin frontend over the same tested engines:

oak --help
oak doctor
oak check
oak replay
oak state inspect --root /path/to/project --slug task-slug

The shell wrappers remain supported. oak check validates the installed target without executing a target-provided checker, while oak replay defaults to the packaged static corpus and synthetic fixtures. See the oak CLI guide for commands, target precedence, exit codes, and explicit exclusions.

oak state runs the packaged durable loop-state runtime against an explicit project root; it is intended for bounded local workflows that need a journal, lock, and recovery without copying harness scripts into the project. Its completion gate requires a durable reviewer subagent APPROVE attestation; see the CLI guide for the exact command.

Preview an ownership-safe upgrade, then apply it:

./upgrade.sh --dry-run
./upgrade.sh

Upgrade aborts before writing if an owned file is missing or modified, an obsolete owned file changed, or an unowned collision exists. Preserved files remain user-owned. When a shipped protected file changes, doctor reports whether the source, user copy, or both changed:

./doctor.sh

Upgrade also compares the local source version with kit_version in the active manifest. A newer source may upgrade; equal version and payload is a no-op; equal version with a different payload is blocked as an identity contradiction; and an older source is blocked to prevent downgrade. doctor reports not-installed, current, upgrade-available, source-older, same-version-different-payload, or invalid-version-state without consulting the network.

Doctor prints twelve ordered findings for OpenCode and Node compatibility, installed dependencies, managed files and drift, required configuration, optional plugins, permissions, executable scripts, the generated skill registry, aggregate compatibility, and legacy residue. Each finding is classified as pass, info, action-required, or not-applicable, followed by a deterministic summary. It does not print configuration contents, subprocess stderr, environment values, credentials, or rollback bytes.

Doctor returns 0 when no finding requires action, 1 for safely actionable state, and 2 for invalid invocation, corrupt/unsafe state, or an unrecoverable filesystem error. Suggested actions are never applied automatically, and preserved custom configuration remains user-owned. See the installation lifecycle guide for the full diagnostic contract. To acknowledge a completed manual merge, run:

./doctor.sh --accept-preserved opencode.json

The command prints a full ACK-PRESERVED line containing the exact target/source hashes and modes. Paste that line back byte-for-byte; yes, a path alone, truncated values, or a stale tuple do not authorize mutation.

Uninstall previews first and removes only unchanged files owned by the manifest. Modified, missing, preserved, and unrelated user files are never deleted:

./uninstall.sh --dry-run
./uninstall.sh
./uninstall.sh --yes

The most recent committed install, upgrade, acknowledgement, or uninstall can be reversed once:

./rollback.sh --dry-run
./rollback.sh

If an operation was interrupted, rollback.sh uses the active journal to restore the pre-operation state instead. Do not delete or edit .oak manually: it contains ownership metadata, recovery state, and potentially sensitive exact backup bytes. See the installation lifecycle guide for command syntax, conflict handling, and recovery details.

Skills

The kit includes local skills that agents use as practical checklists rather than mandatory process overhead.

Included skills cover:

• API and interface design
• Autonomous loops
• Code review and quality
• Code simplification
• Context engineering
• Debugging and error recovery
• Documentation and ADRs
• Doubt-driven development
• Iterative retrieval
• Open Design
• Performance optimization
• Security and hardening
• Source-driven development
• Test-driven development
• Verification loops
• Using agent skills effectively

These skills are available under:

opencode/skills/

They are designed to help agents make better engineering decisions without forcing every task through every checklist.

Optional integrations

The integrations below are optional extensions. They are not required for the core agent workflow.

Integration	What it adds	Setup
Open Design	Editable, local-first design workspace workflows	Open Design guide
Superpowers	Additional skills and software-development workflow discipline	Superpowers guide
Impeccable	Design guidance and deterministic frontend-quality checks	Impeccable guide

Open Design

Open Design is included as an optional local integration for design-oriented workflows.

The designer agent can use:

• Project context from PRODUCT.md and DESIGN.md.
• Optional Impeccable design context.
• Open Design tools through OPEN_DESIGN_URL.

Set OPEN_DESIGN_URL to the base URL of your Open Design workbench:

export OPEN_DESIGN_URL="https://open-design.example.com"

A local or LAN URL is also valid:

export OPEN_DESIGN_URL="http://192.168.1.50:7456"

Do not use a project page or file URL:

# Invalid
export OPEN_DESIGN_URL="https://open-design.example.com/projects/my-project"
export OPEN_DESIGN_URL="https://open-design.example.com/projects/my-project/files/index.html"

Open Design is optional. You can use the rest of the kit without it.

Run Open Design with Docker

A Docker setup is included:

cd docker/open-design
cp .env.example .env
docker compose up -d --build

If Open Design should use OpenCode as a design engine, authenticate inside the container:

docker exec -it open-design bash
opencode auth login
opencode models openai --refresh
exit

Then configure the Open Design base URL:

export OPEN_DESIGN_URL="http://192.168.1.50:7456"

Superpowers

Superpowers is not vendored into this repository.

When plugins are supported and network access is available, the kit can reference the upstream plugin:

{
  "plugin": [
    "superpowers@git+https://github.com/obra/superpowers.git#d884ae04edebef577e82ff7c4e143debd0bbec99"
  ]
}

The human-readable upstream release is v6.1.1; the configuration uses the full reviewed commit because the label alone is not immutable:

{
  "plugin": [
    "superpowers@git+https://github.com/obra/superpowers.git#d884ae04edebef577e82ff7c4e143debd0bbec99"
  ]
}

If your existing opencode.json was preserved during installation, add the plugin manually.

Impeccable

Impeccable is optional and is not included in this repository.

Install it from its upstream source when you want the designer agent to use it as additional design context, particularly when project-specific PRODUCT.md or DESIGN.md files are missing.

Token usage plugin

The bundled TUI plugin can show:

• Lead-agent token usage.
• Total token usage across child and subagent sessions.

This depends on OpenCode exposing the session tree through its TUI plugin API.

If your existing tui.json was preserved during installation, add the bundled plugin manually:

{
  "plugin": [
    "./plugins/token-tree-usage.tsx"
  ]
}

Validation

Normal harness use requires Node.js, npm, and a supported OpenCode version; it does not require a system SQLite client.

Full validation and development additionally require the external sqlite3 binary because the session-evidence tests create real SQLite fixtures. On Ubuntu or Debian, install it with:

sudo apt-get install -y sqlite3

Run the normal local check with:

npm run check

The bundled routing corpus can also be replayed deterministically without calling a model:

node opencode/scripts/replay-routing.mjs \
  --corpus opencode/docs/ai/evolution/benchmarks/router-scenarios.jsonl \
  --fixtures opencode/docs/ai/evolution/benchmarks/replay-fixtures.jsonl

Aggregate one or more sanitized replay reports into deterministic routing metrics without calling a model or rereading raw evidence:

node opencode/scripts/summarize-routing-metrics.mjs \
  --corpus opencode/docs/ai/evolution/benchmarks/router-scenarios.jsonl \
  --report /path/to/routing-replay-report.json

Inconclusive observations are excluded from metric denominators. Token cost is reported as unavailable until the replay report schema carries token evidence.

An isolated live replay is available only through explicit opt-in. It may consume model tokens, is intentionally excluded from normal checks and CI, and does not persist model/provider configuration or raw output by default.

This runs the fast contract checker and every bundled node:test suite. For a fast structural check while editing documentation or contracts, use npm run check:quick. Before a release, use npm run check:release; it performs a frozen dependency install and also runs typechecking, dependency integrity, audit/signature checks, installation smoke, and the exact package smoke. Publication remains separately authorized; follow the reviewed artifact and checksum procedure in the supply-chain policy.

Validate only the canonical identity and current release note with:

npm run check:version
node scripts/version.mjs --check-tag v1.0.39

Tag validation compares an explicitly supplied tag with the package identity. It does not create or push a tag, publish npm content, or create a hosted release. Those remain separate manual, explicitly approved steps.

The contract checker validates the shipped harness, including:

• Configuration JSON.
• Agent and command frontmatter.
• Default lead routing contract.
• /feature sidecar boundaries.
• /plan behavior.
• Agent-readable documentation under opencode/docs/ai/harness/.

The core harness validator is located at:

opencode/scripts/check-harness.mjs

After global installation, you can also run it from the installed OpenCode configuration directory:

node scripts/check-harness.mjs

Project structure

.
├── .github/
│   ├── ISSUE_TEMPLATE/
│   ├── workflows/
│   ├── CODEOWNERS
│   └── PULL_REQUEST_TEMPLATE.md
├── docker/
│   └── open-design/
├── docs/
│   ├── releases/
│   ├── agents.md
│   ├── commands.md
│   ├── configuration.md
│   ├── docker-open-design.md
│   ├── impeccable.md
│   ├── installation.md
│   ├── models.md
│   ├── open-design.md
│   ├── quickstart.md
│   ├── security.md
│   ├── superpowers.md
│   ├── synology.md
│   ├── troubleshooting.md
│   └── workflows.md
├── opencode/
│   ├── agents/
│   ├── commands/
│   ├── docs/ai/
│   ├── plugins/
│   ├── references/
│   ├── scripts/
│   ├── skills/
│   ├── tools/
│   ├── AGENTS.md
│   ├── opencode.json
│   └── tui.json
├── scripts/
│   ├── check.sh
│   ├── install-smoke.sh
│   ├── manage-installation.mjs
│   └── manage-installation.test.mjs
├── doctor.sh
├── install.sh
├── rollback.sh
├── uninstall.sh
├── upgrade.sh
└── env.example

The shipped OpenCode configuration contains:

Path	Purpose
AGENTS.md	Global behavior rules and agent index
opencode.json	Models, permissions, plugins, and default agent
tui.json	TUI plugin registration
agents/	Specialized agent prompts
commands/	Slash-command workflows
skills/	Local process skills
tools/	Custom TypeScript tools
plugins/	Bundled OpenCode/TUI plugins
references/	Reusable checklists referenced by skills
docs/ai/harness/	Agent, command, evidence, and validation contracts
docs/ai/evolution/	Harness-evolution benchmark and evidence records
scripts/check-harness.mjs	Mechanical harness validation

Permissions and safety

The default OpenCode permissions are conservative:

• Reads are allowed.
• Edits ask by default.
• Bash commands ask by default.
• External directories are denied.

Security recommendations:

• Do not commit .env, authentication files, sessions, logs, or provider credentials.
• Do not commit private PRODUCT.md or DESIGN.md files unless intended.
• Treat Open Design as a privileged local tool because it can run local agent CLIs and write files inside project workspaces.
• Do not expose Open Design directly to the Internet without authentication.
• Prefer localhost, LAN, VPN, Tailscale, WireGuard, or authenticated HTTPS through a reverse proxy.

Troubleshooting

OPEN_DESIGN_URL is not set

Set the base URL only:

export OPEN_DESIGN_URL="http://192.168.1.50:7456"

Do not use a project-specific path.

/api/health fails

Check that Open Design is running and reachable from the machine or container where OpenCode runs.

OpenCode does not appear in Open Design agents

Run:

opencode auth login

Also verify that opencode is available on PATH.

Superpowers skills do not load

Restart OpenCode and verify that the Superpowers plugin entry is present in opencode.json.

Token usage does not appear in the TUI

Verify:

• tui.json contains the bundled plugin entry.
• npm ci --ignore-scripts was run in the OpenCode config directory while its shipped manifests were intact (or preserved manifests were reviewed and merged before selecting another install command).
• OpenCode was restarted.
• The current OpenCode TUI exposes child-session information to plugins.

Designer cannot access Open Design

Verify:

• OPEN_DESIGN_URL.
• Tool registration.
• opencode/tools/open_design.ts.
• Network reachability between OpenCode and Open Design.

crypto.randomUUID fails over HTTP on a LAN URL

Use HTTPS, or apply the optional upstream frontend patch described in the documentation.

Documentation

Detailed guides are available in docs/:

• Quickstart
• Installation
• Configuration
• Agents
• Commands
• Workflows
• Models
• Open Design
• Docker Open Design
• Superpowers
• Impeccable
• Security
• Threat model
• Reproducible use cases
• Troubleshooting
• Synology notes

Compatibility and scope

• Node.js 22 and 24 are supported; the exact engine is ^22.9.0 || ^24.0.0.
• OpenCode >=1.14.41 <2.0.0 is supported through tested boundary versions.
• WSL2 is experimental. Native Windows is unsupported because lifecycle wrappers require Bash.
• Open Design, Superpowers, Impeccable, Docker, and token-usage plugins remain optional.

See the complete compatibility matrix for status definitions, exact pins, evidence, and scope.

Contributing

Contributions are welcome.

Found a bug or have an idea? Open an issue · Browse existing issues

Good contributions include:

• Clearer documentation.
• Safer installation or uninstall behavior.
• Improved troubleshooting.
• Tighter agent or command prompts.
• Better Docker and Open Design setup.
• Validation improvements.
• Reusable process skills.
• Focused workflow enhancements that remain portable and safe for public reuse.

Before opening a pull request:

npm run check

Use npm run check:release for release readiness.

If Docker files changed:

docker compose -f docker/open-design/docker-compose.yml config

Please avoid committing credentials, authentication files, sessions, logs, private product documents, or machine-specific paths.

See CONTRIBUTING.md for details.

License

Licensed under the Apache License 2.0.

See NOTICE.md for attribution notices.

Disclaimer

This repository is not affiliated with OpenCode, Open Design, Impeccable, or Superpowers.