|
|
@@ -212,12 +212,115 @@ If any agent fails to respond, check your provider authentication and config fil
|
|
|
|
|
|
### What's New in V2
|
|
|
|
|
|
-V2 is a scheduler-first multi-agent workflow: the Orchestrator plans,
|
|
|
-delegates, reconciles results, and verifies outcomes while specialists work in
|
|
|
-their own lanes. Start with the [OpenAI GPT-5.6 spotlight](#v2-is-ready-for-openai-gpt-56),
|
|
|
-then see [Background Orchestration](docs/background-orchestration.md),
|
|
|
-[Companion](docs/companion.md), and [Skills](docs/skills.md) for Deepwork,
|
|
|
-Reflect, Worktrees, and the bundled configuration skill.
|
|
|
+V2 turns oh-my-opencode-slim into a scheduler-first multi-agent workflow system.
|
|
|
+The Orchestrator stays focused on planning, delegation, reconciliation, and
|
|
|
+verification while specialists do the work in their own lanes.
|
|
|
+
|
|
|
+- **[Background agents](#background-agents)** - the Orchestrator now dispatches
|
|
|
+ specialists as background tasks, tracks task/session IDs, waits for completion
|
|
|
+ events, and reconciles results before continuing.
|
|
|
+- **[Companion](#companion)** - an optional floating desktop window shows which
|
|
|
+ agents are currently active, including parallel background specialists.
|
|
|
+- **[Deepwork](#deepwork)** - a structured workflow for large, multi-file, risky,
|
|
|
+ or phased coding work using persistent plan files and Oracle review gates.
|
|
|
+- **[Reflect](#reflect)** - reviews repeated work patterns and suggests reusable skills,
|
|
|
+ agents, commands, config rules, prompt rules, or project playbooks.
|
|
|
+- **[Worktrees](#worktrees)** - manages Git worktrees as isolated coding lanes
|
|
|
+ with safety protocols for complex, risky, or parallel tasks.
|
|
|
+- **[oh-my-opencode-slim skill](#oh-my-opencode-slim-skill)** - a bundled
|
|
|
+ configuration skill that helps tune models, prompts, custom agents, MCP access,
|
|
|
+ presets, and plugin behavior safely.
|
|
|
+
|
|
|
+#### Background Agents
|
|
|
+
|
|
|
+V2 makes background specialists the default mental model: the Orchestrator plans
|
|
|
+the work graph, launches the right agents, avoids overlapping write ownership,
|
|
|
+and waits for terminal task results before acting on them.
|
|
|
+
|
|
|
+See **[Background Orchestration](docs/background-orchestration.md)** for the
|
|
|
+full scheduler model.
|
|
|
+
|
|
|
+#### Companion
|
|
|
+
|
|
|
+The optional Companion is a floating desktop status window for live agent
|
|
|
+activity. It shows the current session state and which agents are active, so
|
|
|
+background work is easier to follow at a glance.
|
|
|
+
|
|
|
+<div align="center">
|
|
|
+ <img src="img/companion.gif" alt="Companion showing active agents" width="600">
|
|
|
+ <p><i>Left bottom visual companion.</i></p>
|
|
|
+</div>
|
|
|
+
|
|
|
+During interactive install, the installer asks whether to enable Companion and
|
|
|
+defaults to `no`. For automation, enable it explicitly with:
|
|
|
+
|
|
|
+```bash
|
|
|
+bunx oh-my-opencode-slim@latest install --companion=yes
|
|
|
+```
|
|
|
+
|
|
|
+See **[Companion](docs/companion.md)** for configuration, positions, sizes, and
|
|
|
+install details.
|
|
|
+
|
|
|
+#### Deepwork
|
|
|
+
|
|
|
+Deepwork is for heavy coding sessions: broad refactors, multi-phase features,
|
|
|
+risky architecture changes, or work that needs a persistent plan. It creates a
|
|
|
+local markdown progress file, uses Oracle review gates, and keeps implementation
|
|
|
+phases structured.
|
|
|
+
|
|
|
+Start it with:
|
|
|
+
|
|
|
+```text
|
|
|
+/deepwork <heavy coding task>
|
|
|
+```
|
|
|
+
|
|
|
+See **[Skills](docs/skills.md#deepwork)** for when to use it and how the workflow
|
|
|
+runs.
|
|
|
+
|
|
|
+#### Reflect
|
|
|
+
|
|
|
+Reflect helps the Orchestrator learn from repeated workflow friction. It reviews
|
|
|
+recent work and existing assets, then recommends the smallest useful improvement:
|
|
|
+a skill, custom agent, command, config rule, prompt rule, MCP permission change,
|
|
|
+or project playbook. If there is not enough evidence, it should recommend
|
|
|
+creating nothing.
|
|
|
+
|
|
|
+Use it directly with:
|
|
|
+
|
|
|
+```text
|
|
|
+/reflect
|
|
|
+/reflect release workflow and checks
|
|
|
+```
|
|
|
+
|
|
|
+Or with natural prompts like:
|
|
|
+
|
|
|
+```text
|
|
|
+reflect on my recent workflows
|
|
|
+find repeated work worth turning into reusable instructions
|
|
|
+```
|
|
|
+
|
|
|
+See **[Skills](docs/skills.md#reflect)** for the full workflow and guardrails.
|
|
|
+
|
|
|
+#### Worktrees
|
|
|
+
|
|
|
+Worktrees manages Git worktrees as safe, isolated coding lanes under `.slim/worktrees/<slug>/`. The Orchestrator manages the lifecycle of these lanes, tracks state in `.slim/worktrees.json`, dispatches specialist agents inside them, and requires explicit confirmation before mutating git state.
|
|
|
+
|
|
|
+See **[Skills](docs/skills.md#worktrees)** for the safety protocol.
|
|
|
+
|
|
|
+#### oh-my-opencode-slim Skill
|
|
|
+
|
|
|
+The bundled `oh-my-opencode-slim` skill helps the Orchestrator configure and
|
|
|
+improve the plugin itself. Use it for model tuning, custom agents, prompt
|
|
|
+overrides, skill/MCP permissions, presets, optional agents, background
|
|
|
+orchestration, and recurring workflow friction.
|
|
|
+
|
|
|
+<div align="center">
|
|
|
+ <img src="img/oh-my-opencode-skill.png" alt="oh-my-opencode-slim skill in use" width="600">
|
|
|
+ <p><i>Ask the bundled skill to tune and improve your agent setup.</i></p>
|
|
|
+</div>
|
|
|
+
|
|
|
+See **[Skills](docs/skills.md#oh-my-opencode-slim)** for examples and safety
|
|
|
+rules.
|
|
|
|
|
|
---
|
|
|
|
Alvin Unreal