├── web/ # Marketing site (Next.js 15 + Cloudflare Workers, deployed to ohmyopenagent.com via opennextjs-cloudflare). Independent package with own bun.lock — see web/AGENTS.md
OFF by default. Parallel multi-agent coordination, modeled after Claude Code Agent Teams. Enable via `team_mode.enabled` in `.opencode/oh-my-opencode.jsonc` or user config; restart OpenCode after change.
Teams live as directories under `~/.omo/teams/{name}/config.json` (user) or `<project>/.omo/teams/{name}/config.json` (project; project beats user on collisions). Members declared as `kind: "subagent_type"` (direct agent) or `kind: "category"` (routed through `sisyphus-junior`).
-`conditional`: hephaestus (lacks `teammate: "allow"` permission by default — apply D-36 in `tool-config-handler.ts` or use `subagent_type: "sisyphus"` instead)
-`hard-reject`: oracle, librarian, explore, multimodal-looker, metis, momus, prometheus (rejected at parse — use `task`/delegate-task)
**Implementation:** [`src/features/team-mode/`](file:///Users/yeongyu/local-workspaces/omo/src/features/team-mode/AGENTS.md). User docs: [`docs/guide/team-mode.md`](file:///Users/yeongyu/local-workspaces/omo/docs/guide/team-mode.md).
| Add new hook | `src/hooks/{name}/` + register in `src/plugin/hooks/create-*-hooks.ts` | Pick the right tier (Session/ToolGuard/Transform/Continuation/Skill) |
| Add new tool | `src/tools/{name}/` + register in `src/plugin/tool-registry.ts` | Factory `createXXXTool` (most) or direct `ToolDefinition` (LSP, interactive_bash) |
| Add new feature module | `src/features/{name}/` | Standalone module wired into `plugin/` layer |
| Add new MCP (tier 1) | `src/mcp/` + register in `createBuiltinMcps()` | Remote HTTP only |
| Add new built-in skill | `src/features/builtin-skills/skills/{name}.ts` + register in `skills.ts` | Implement `BuiltinSkill` interface |
| Add new command | `src/features/builtin-commands/` | Templates in `templates/` |
- **Canonical agent order:** Sisyphus → Hephaestus → Prometheus → Atlas. Enforced by `installAgentSortShim()` (patches `Array.prototype.toSorted`/`.sort` narrowly when the array contains ≥2 canonical core agents). See [`src/plugin-handlers/AGENTS.md`](file:///Users/yeongyu/local-workspaces/omo/src/plugin-handlers/AGENTS.md) for the full history of why this exists.
- **Hashline edit + read pairing:** Every `Read` tool output is tagged with `LINE#ID` content hashes; `hashline_edit` validates the hash before applying. Stale hash → reject.
- **Per-session MCP isolation:** Tier-3 MCP clients keyed by `${sessionID}:${skillName}:${serverName}` so the same skill in two sessions does not share state.
- **Two fallback systems:** `model-fallback` (proactive, chat.params) vs `runtime-fallback` (reactive, session.error). They operate independently — no direct integration.
- **OpenClaw bidirectional:** Outbound dispatchers fire on session events; inbound daemon polls Discord/Telegram and `send-keys` replies into the tracked tmux pane.
- **Internal message injection is dangerous:** OpenCode의 stupid한 설계로 플러그인이 `session.prompt` / `session.promptAsync` 같은 메인 세션 메시지 API를 통해 메인 시스템을 망가뜨릴 수 있다. Any route that injects a message into an existing session must go through `src/shared/prompt-async-gate.ts` or an equivalent per-session reservation/status gate, must check active session state before dispatch, and must have a regression test proving duplicate concurrent injections collapse to one dispatch. This applies to background completion wakes, fallback retries, team mailbox live delivery, recovery continuations, CLI run resumes, Claude Code hook injections, and sync/background subagent prompts.
- **Dual package:** `oh-my-opencode` + `oh-my-openagent` published simultaneously during the rename transition.
- **Comments:** AI slop comment patterns blocked by `comment-checker` hook (binary: `@code-yeongyu/comment-checker`). Use `// @allow` to bypass single line, `// comment-checker-disable-file` at file top to bypass file. Sparingly.
## ANTI-PATTERNS (BLOCKING)
- Never `as any`, `@ts-ignore`, `@ts-expect-error`.
- Never suppress lint/type errors.
- Never add emojis to code/comments unless user explicitly asks.
- Never commit unless explicitly requested.
- Never run `bun publish` directly — use the GitHub Actions workflow.
- Never modify `package.json``version` locally — handled by publish workflow.
- Never write to existing files without reading them first (`write-existing-file-guard`).
- Never use `background_cancel(all=true)` — cancel by `taskId` individually.
- Never delete a failing test to make a build green. Fix the code.
- Never em dashes / en dashes / AI filler ("simply", "obviously", "clearly", "moreover", "furthermore") in generated content.
- Never create catch-all files (`utils.ts`, `helpers.ts`, `service.ts`).
- Never empty catch blocks `catch(e) {}`.
- Never test with Arrange-Act-Assert comments — use given/when/then.
- Never dump business logic into `index.ts` — barrel exports only.
- Prometheus may ONLY edit `.md` files (enforced by `prometheus-md-only` hook); FORBIDDEN paths: `src/`, `package.json`, config files.
- **Logger:** writes to `/tmp/oh-my-opencode.log` — check there for debugging.
- **Background tasks:** 5 concurrent per `${providerID}/${modelID}` key by default (configurable via `background_task.modelConcurrency` / `providerConcurrency`); FIFO queue when slots full.
- **Plugin load timeout:** 10s for Claude Code plugin discovery.
- **Model fallback:** per-agent chains in `src/shared/model-requirements.ts`. **There is no single global priority.**
- **CI test isolation:** `script/run-ci-tests.ts` auto-isolates files using `mock.module()` (plus `src/openclaw/__tests__/reply-listener-discord.test.ts`) — they run in separate processes.
- **Architecture rules** enforced via `.sisyphus/rules/modular-code-enforcement.md` (when present in workspace).
- **Windows builds:** run on `windows-latest` (not cross-compiled) to avoid Bun segfaults.
- **Platform binaries:** detect AVX2 + libc family at runtime, fallback to baseline if needed.
- **IntentGate (`keyword-detector`):** classifies user intent (`ultrawork`/`ulw`, `search`, `analyze`, `team`) and injects mode-specific prompts.
- **Hashline edit:** every `Read` output tagged with `LINE#ID` content hashes (chars from `ZPMQVRWSNKTXJBYH`); edits reject on hash mismatch.
- **Docs:** see [`docs/guide/`](file:///Users/yeongyu/local-workspaces/omo/docs/guide/) for user-facing guides (overview, installation, orchestration, agent-model-matching, team-mode), [`docs/reference/`](file:///Users/yeongyu/local-workspaces/omo/docs/reference/) for CLI/configuration/features reference.
