Oh My OpenCode Features
Overview
This skill provides a comprehensive reference of all features available in oh-my-opencode, including curated agents, powerful tools, MCP integrations, hook systems, and productivity enhancements.
When to Use
Trigger phrases:
"oh my opencode features"
"Agent orchestration – When you need specialized agents for complex tasks, de"
"MultiModel workflows – When leveraging multiple LLM providers, fallback chai"
"Background execution – When running parallel agents, async task completion,"
Agent orchestration – When you need specialized agents for complex tasks, deep work, architecture decisions, documentation, or codebase exploration
MultiModel workflows – When leveraging multiple LLM providers, fallback chains, or per-agent model optimization
Background execution – When running parallel agents, async task completion, or autonomous deep work
LSP/AST tools – When refactoring, renaming, searching code with AST awareness, or running diagnostics
Hooks integration – When automating workflows via PreToolUse, PostToolUse, UserPromptSubmit, Stop triggers
Session management – When managing session history, session continuity, or session-based task continuation
Workflow automation – When using Ultrawork, Ralph Loop, TDD, debugging patterns
Quality verification – When enforcing comment checks, TODO enforcement, and verification before completion
Core Features
- Core operation execution with comprehensive error handling
- Input validation and output quality assurance
- Integration with existing workflows and toolchains
- Detailed logging for debugging and audit trails
1. Curated Agent System
Sisyphus — Main Orchestrator
- Purpose: Primary autonomous agent for complex tasks
- Key Features:
- Goal-oriented execution without step-by-step instructions
- Multi-model orchestration
- Parallel agent launches (2-5 agents simultaneously)
- Strict TODO list enforcement
- Hash-anchored edit tool (LINE#ID format)
- Comment checker for clean code
- Claude Code compatibility layer
Hephaestus — Autonomous Deep Worker
- Purpose: Goal-oriented autonomous problem-solving
- Key Features:
- Explores before acting (2-5 parallel explore/librarian agents)
- End-to-end completion with verification
- Pattern matching for codebase consistency
- Surgical, minimal code generation
- Thorough research before action
Oracle — Architecture & Debugging
- Purpose: High-IQ strategic consultation
- Key Features:
- Complex debugging after 2+ failed attempts
- Architecture design decisions
- Multi-system tradeoffs analysis
- Security/performance concerns review
- Read-only consultation model
Librarian — Documentation & Search
- Purpose: External knowledge retrieval
- Key Features:
- Official documentation search
- Open source implementation patterns
- Real-time source code digestion
- Best practices for unfamiliar libraries
- Multi-source search (docs, GitHub, Stack Overflow)
Explore — Fast Codebase Exploration
- Purpose: Rapid codebase analysis
- Key Features:
- Blazing fast contextual grep
- Pattern discovery across modules
- Multiple search angles simultaneously
- Quick file and symbol location
2. Multi-Model Orchestration
Model Selection
- Support for multiple LLM providers:
- OpenAI (GPT-4o, GPT-4 Turbo, etc.)
- Anthropic (Claude 3.5, Claude 3, etc.)
- Google (Gemini Pro, etc.)
- OpenCode Zen (curated models)
Fallback Chains
- Automatic fallback on model failure
- Configurable fallback order
- Per-agent model selection
Model Optimization
- Temperature tuning per agent
- Context window optimization
- Token usage tracking
3. Background Task Execution
Parallel Execution
- Run multiple agents simultaneously
- Async task completion
- Concurrent tool execution
- Background task persistence
Task Management
- Task queue with prioritization
- Progress tracking and reporting
- Task cancellation and restart
- Background/foreground switching
4. LSP & AST Tools
Built-in LSP Tools
- lsp_diagnostics: Get errors, warnings, hints
- lsp_goto_definition: Jump to symbol definition
- lsp_find_references: Find all usages of symbol
- lsp_rename: Rename symbol across workspace
- lsp_symbols: Get symbols from file/workspace
- lsp_prepare_rename: Check if rename is valid
AST-Grep Tools
- ast_grep_search: AST-aware code search
- ast_grep_replace: AST-aware code replacement
- Multi-language support (25+ languages)
Refactoring Capabilities
- Deterministic, surgical refactoring
- Safe multi-file changes
- Pattern-based code transformation
5. Hash-Anchored Edit Tool
How It Works
- Uses
LINE#ID format for edits
- Validates content hash before applying changes
- Prevents stale-line edit errors
- Atomic, verifiable modifications
Usage Pattern
// Before: Line 42 contains specific content
// Edit is anchored to content hash, not line number
// Ensures edit applies to intended location
6. MCP (Model Context Protocol) Integration
Built-in MCPs
| MCP |
Provider |
Purpose |
| websearch |
Exa |
Web search for current information |
| context7 |
Context7 |
Official documentation lookup |
| grep_app |
Grep.app |
GitHub code search |
| playwright |
Browser |
Browser automation |
| git-master |
Git |
Git operations |
Custom MCPs
- Add custom MCP servers
- Configure MCP groups
- Per-project MCP settings
- MCP server authentication
7. Hook System
Available Hooks
| Hook |
Purpose |
Usage |
| PreToolUse |
Validate before tool execution |
Security, logging |
| PostToolUse |
Process after tool execution |
Verification, cleanup |
| UserPromptSubmit |
Process user prompts |
Context injection |
| Stop |
Cleanup on session end |
Resource cleanup |
| PreAgentRun |
Before agent execution |
Setup, validation |
| PostAgentRun |
After agent completion |
Review, logging |
Hook Configuration
- Enable/disable specific hooks
- Chain multiple handlers
- Custom hook scripts
- Per-agent hook settings
8. Claude Code Compatibility
Supported Features
- Commands:
/ commands for common actions
- Agents: Full agent system compatibility
- Skills: Skill tool integration
- MCP: MCP server integration
- Hooks: PreToolUse, PostToolUse, UserPromptSubmit, Stop
Compatibility Mode
- Seamless migration from Claude Code
- Identical command interface
- Shared configuration patterns
9. Session Management
Session Features
- Session List: View all sessions
- Session Read: Review previous sessions
- Session Search: Find by content
- Session Export: Share or backup
Session Persistence
- Configurable storage location
- Automatic session saving
- Session continuity across restarts
- Max session count management
10. Todo System
Features
- TODO List: Track all tasks
- TODO Write: Create/manage todos
- Status Tracking: Pending, in_progress, completed
- Auto-Continuation: Forces agent to finish
Enforcement
- Strict TODO following
- Auto-reminder for incomplete tasks
- Task completion verification
- Progress reporting
Productivity Features
- Core operation execution with comprehensive error handling
- Input validation and output quality assurance
- Integration with existing workflows and toolchains
- Detailed logging for debugging and audit trails
1. Ralph Loop
Purpose: Self-referential development loop
Features:
- Automatic plan execution and validation
- Iterative improvement
- Goal tracking
- Progress monitoring
2. Ultrawork Mode
Activation: Include "ultrawork" or "ulw" in prompt
Effects:
- Parallel agents activate
- Background tasks enabled
- Deep exploration mode
- Relentless execution
3. Comment Checker
Purpose: Maintain code quality
Rules:
- Prevents excessive comments
- Justifies necessary comments
- Enforces code readability
- Follows project style
4. Context Injection
Features:
- Auto-inject AGENTS.md
- Conditional rules injection
- Project context awareness
- User preferences preservation
5. Auto-Completion
Features:
- End-to-end task completion
- Verification before finishing
- Evidence-based reporting
- Quality gate enforcement
Workflow Enhancements
- Receive input and validate format
- Route to appropriate handler based on input type
- Execute core operation with monitoring
- Transform output to expected format
- Return results or trigger follow-up actions
1. Parallel Agent Dispatching
// Fire multiple agents for independent tasks
- Explore: Map codebase structure
- Librarian: Find relevant documentation
- Hephaestus: Research best practices
- Oracle: Review architecture
2. Subagent-Driven Development
// Dispatch fresh subagent per task
- Two-stage review (spec, quality)
- No context pollution
- Parallel-safe execution
- Review checkpoints
3. Git Worktree Integration
// Isolated workspace per task
- Create worktree for feature
- Commit to feature branch
- Merge when complete
- Clean up when done
4. Test-Driven Development
// Subagents follow TDD
- Write test first
- Implement feature
- Verify tests pass
- Refactor as needed
5. Systematic Debugging
// Structured debugging approach
- Reproduce issue
- Identify root cause
- Implement fix
- Verify solution
- Document learnings
Tool Categories
| Tool |
Purpose |
Required |
| CLI |
Primary execution |
Yes |
| API client |
External service calls |
Conditional |
| Validator |
Output checking |
Recommended |
File Operations
- read: Read file contents
- write: Write/overwrite files
- edit: Modify file contents
- glob: Find files by pattern
- grep: Search file contents
Code Analysis
- lsp_diagnostics: Get diagnostics
- lsp_goto_definition: Jump to definition
- lsp_find_references: Find references
- lsp_rename: Rename symbol
- ast_grep_search: AST search
- ast_grep_replace: AST replace
System Operations
- bash: Execute shell commands
- interactive_bash: Interactive terminal
- session_*: Session management
- Notion_*: Notion integration
Browser Operations
- skill_mcp: MCP server tools
- playwright: Browser automation
- dev-browser: Persistent browser state
- agent-browser: CLI browser automation
- browser-use: Web automation
Web Operations
- websearch_web_search_exa: Web search
- google_search: Google search + analysis
- webfetch: Fetch URL content
- look_at: Analyze media files
Configuration Categories
{
"enabled": true,
"autoRun": false,
"timeout": 30000,
"retries": 3
}
Set environment variables as needed for authentication and endpoints.
Agent Configuration
- Model selection per agent
- Temperature tuning
- Permission settings
- Custom prompts
Feature Configuration
- Parallel execution limits
- Background task settings
- Hook enable/disable
- Quality gate settings
Integration Configuration
- MCP server setup
- Hook script configuration
- Plugin integration
- External tool connection
UI Configuration
- Theme selection
- Keybind customization
- Layout preferences
- Display options
Security Features
- Core operation execution with comprehensive error handling
- Input validation and output quality assurance
- Integration with existing workflows and toolchains
- Detailed logging for debugging and audit trails
Permission System
- File system access control
- Network access restrictions
- Shell command allowlisting
- API key management
Validation
- Input validation before execution
- Content hash verification
- Safe mode options
- Audit logging
Compliance
- ToS-aware configuration
- Security best practices
- Audit trail maintenance
- Secret management
Performance Features
- Core operation execution with comprehensive error handling
- Input validation and output quality assurance
- Integration with existing workflows and toolchains
- Detailed logging for debugging and audit trails
Optimization
- Intelligent context truncation
- Token usage tracking
- Parallel execution efficiency
- Cache management
Scalability
- Concurrent agent limits
- Resource usage monitoring
- Memory optimization
- Background task queuing
Extensibility
This section covers extensibility for the oh-my-opencode-features skill.
Key operations include input validation, core processing, and output verification.
Refer to the skill overview for detailed usage instructions.
Custom Tools
- Create custom tools
- Define tool schemas
- Implement execution logic
- Register with system
Custom Hooks
- Write custom hooks
- Script integration
- Event-driven automation
- Workflow customization
Plugin System
- Extend functionality
- Add new agents
- Integrate external services
- Customize behavior
Comparison with Other Harnesses
| Feature |
Oh My OpenCode |
Standard OpenCode |
Claude Code |
| Agents |
5 curated |
Basic |
1 |
| Parallel Tasks |
Yes |
Limited |
No |
| Background Exec |
Yes |
No |
No |
| Hook System |
25+ |
Basic |
4 |
| MCP Built-ins |
5+ |
0 |
0 |
| Claude Compat |
Full |
Full |
Native |
| Hash Anchoring |
Yes |
No |
No |
| TODO Enforcement |
Yes |
No |
No |
Feature Quick Reference
| Category |
Key Features |
| Agents |
Sisyphus, Hephaestus, Oracle, Librarian, Explore |
| Models |
OpenAI, Anthropic, Google, OpenCode Zen |
| Tools |
LSP, AST-Grep, File Ops, Bash, Session |
| MCPs |
websearch, context7, grep_app, playwright, git-master |
| Hooks |
25+ hooks, custom scripts, per-agent settings |
| Sessions |
List, Read, Search, Export, Persistence |
| Workflow |
Ultrawork, Ralph Loop, TDD, Debugging |
| Quality |
Comment Checker, TODO Enforcement, Verification |
The Process
- Identify feature needs – Determine which oh-my-opencode features address your task (agents, MCPs, hooks, sessions, workflows)
- Verify feature availability – Check feature is enabled in configuration and compatible with your OpenCode version
- Configure feature – Set up MCP servers, hooks, categories as needed in config
- Load feature in session – Ensure feature is accessible (agents via /agent, MCPs via /mcp, etc.)
- Use feature – Execute feature-specific workflows (parallel agents, hooks, ultrawork, etc.)
- Monitor execution – Watch feature behavior, adjust configs, debug issues
- Verify output – Run lsp_diagnostics, build, tests to confirm feature completed correctly
When NOT to Use
- Task is outside your authorization scope
- You need to implement controls (use implementing-* skills)
- Task is about analysis, not action (use analyzing-* skills)
- You don't have access to target systems
- Task requires compliance expertise (consult professionals)
- Task is about defense, not offense (use defensive skills)
Red Flags
- Feature conflicts – Enabling multiple conflicting features (e.g., multiple LSP tools fighting over same file)
- Version mismatch – Using features from different oh-my-opencode versions causing incompatibilities
- Missing dependencies – Enabling MCPs without required tools or permissions
- Hook loops – Hooks triggering themselves infinitely due to improper conditions
- Over-enforcement – Enabling all quality checks when only specific ones needed causes slowdowns
- Ignoring workflows – Not following Ultrawork, Ralph Loop, or TDD when they should apply
- Session corruption – Attempting to use features on stale or corrupted sessions
Verification
- Feature loads: Feature commands work (e.g.,
/mcp list, /agent sisyphus)
- MCPs respond: MCP servers execute tool calls and return expected results
- Hooks fire: Hook conditions trigger expected actions at right times
- Sessions valid: Session commands return valid data without errors
- Workflows execute: Ultrawork completes autonomously, TDD creates tests, Ralph Loop iterates to completion
- Quality checks pass: Comment checker, TODO enforcement, verification run as configured
- Integration clean: All features work together without conflicts or race conditions
References
Anti-Rationalization Table
| Rationalization |
Reality |
| "I will handle auth later" |
Retrofitting auth is 10x harder. Build it from day one. |
| "APIs do not change" |
APIs change. Version your integrations and handle deprecations. |
| "Webhooks are optional" |
Without webhooks, you miss real-time events. They are essential. |
Related Skills
- oh-my-opencode: Overall integration
- oh-my-opencode-installation: Setup guide
- oh-my-opencode-agents: Agent deep dive
- oh-my-opencode-usage: Daily usage
- oh-my-opencode-configuration: Advanced setup
Process
- Analyze the task requirements
- Apply domain expertise
- Verify output quality
1---2name: oh-my-opencode-features3description: Use when complete reference of all oh-my-opencode features including4 agents, tools, MCPs, hooks, workflow automation, and productivity enhancements.5 Use when working with oh my opencode features.6domain: integrations7author: oyi778license: Apache-2.09subdomain: integrations10tags:11- ai-agent12- api13- features14- integrations15- opencode16- third-party17- workflow18version: 1.0.019category: integrations20metadata:21 source: https://skillmd.com/skills/oyi77/oh-my-opencode-features22---232425# Oh My OpenCode Features2627## Overview2829This skill provides a comprehensive reference of all features available in oh-my-opencode, including curated agents, powerful tools, MCP integrations, hook systems, and productivity enhancements.3031## When to Use3233**Trigger phrases:**34- "oh my opencode features"35- "**Agent orchestration** – When you need specialized agents for complex tasks, de"36- "**MultiModel workflows** – When leveraging multiple LLM providers, fallback chai"37- "**Background execution** – When running parallel agents, async task completion,"383940- **Agent orchestration** – When you need specialized agents for complex tasks, deep work, architecture decisions, documentation, or codebase exploration41- **MultiModel workflows** – When leveraging multiple LLM providers, fallback chains, or per-agent model optimization42- **Background execution** – When running parallel agents, async task completion, or autonomous deep work43- **LSP/AST tools** – When refactoring, renaming, searching code with AST awareness, or running diagnostics44- **Hooks integration** – When automating workflows via PreToolUse, PostToolUse, UserPromptSubmit, Stop triggers45- **Session management** – When managing session history, session continuity, or session-based task continuation46- **Workflow automation** – When using Ultrawork, Ralph Loop, TDD, debugging patterns47- **Quality verification** – When enforcing comment checks, TODO enforcement, and verification before completion4849## Core Features50- Core operation execution with comprehensive error handling51- Input validation and output quality assurance52- Integration with existing workflows and toolchains53- Detailed logging for debugging and audit trails545556### 1. Curated Agent System5758#### Sisyphus — Main Orchestrator59- **Purpose**: Primary autonomous agent for complex tasks60- **Key Features**:61 - Goal-oriented execution without step-by-step instructions62 - Multi-model orchestration63 - Parallel agent launches (2-5 agents simultaneously)64 - Strict TODO list enforcement65 - Hash-anchored edit tool (LINE#ID format)66 - Comment checker for clean code67 - Claude Code compatibility layer6869#### Hephaestus — Autonomous Deep Worker70- **Purpose**: Goal-oriented autonomous problem-solving71- **Key Features**:72 - Explores before acting (2-5 parallel explore/librarian agents)73 - End-to-end completion with verification74 - Pattern matching for codebase consistency75 - Surgical, minimal code generation76 - Thorough research before action7778#### Oracle — Architecture & Debugging79- **Purpose**: High-IQ strategic consultation80- **Key Features**:81 - Complex debugging after 2+ failed attempts82 - Architecture design decisions83 - Multi-system tradeoffs analysis84 - Security/performance concerns review85 - Read-only consultation model8687#### Librarian — Documentation & Search88- **Purpose**: External knowledge retrieval89- **Key Features**:90 - Official documentation search91 - Open source implementation patterns92 - Real-time source code digestion93 - Best practices for unfamiliar libraries94 - Multi-source search (docs, GitHub, Stack Overflow)9596#### Explore — Fast Codebase Exploration97- **Purpose**: Rapid codebase analysis98- **Key Features**:99 - Blazing fast contextual grep100 - Pattern discovery across modules101 - Multiple search angles simultaneously102 - Quick file and symbol location103104### 2. Multi-Model Orchestration105106#### Model Selection107- Support for multiple LLM providers:108 - OpenAI (GPT-4o, GPT-4 Turbo, etc.)109 - Anthropic (Claude 3.5, Claude 3, etc.)110 - Google (Gemini Pro, etc.)111 - OpenCode Zen (curated models)112113#### Fallback Chains114- Automatic fallback on model failure115- Configurable fallback order116- Per-agent model selection117118#### Model Optimization119- Temperature tuning per agent120- Context window optimization121- Token usage tracking122123### 3. Background Task Execution124125#### Parallel Execution126- Run multiple agents simultaneously127- Async task completion128- Concurrent tool execution129- Background task persistence130131#### Task Management132- Task queue with prioritization133- Progress tracking and reporting134- Task cancellation and restart135- Background/foreground switching136137### 4. LSP & AST Tools138139#### Built-in LSP Tools140- **lsp_diagnostics**: Get errors, warnings, hints141- **lsp_goto_definition**: Jump to symbol definition142- **lsp_find_references**: Find all usages of symbol143- **lsp_rename**: Rename symbol across workspace144- **lsp_symbols**: Get symbols from file/workspace145- **lsp_prepare_rename**: Check if rename is valid146147#### AST-Grep Tools148- **ast_grep_search**: AST-aware code search149- **ast_grep_replace**: AST-aware code replacement150- Multi-language support (25+ languages)151152#### Refactoring Capabilities153- Deterministic, surgical refactoring154- Safe multi-file changes155- Pattern-based code transformation156157### 5. Hash-Anchored Edit Tool158159#### How It Works160- Uses `LINE#ID` format for edits161- Validates content hash before applying changes162- Prevents stale-line edit errors163- Atomic, verifiable modifications164165#### Usage Pattern166```typescript167// Before: Line 42 contains specific content168// Edit is anchored to content hash, not line number169// Ensures edit applies to intended location170```171172### 6. MCP (Model Context Protocol) Integration173174#### Built-in MCPs175176| MCP | Provider | Purpose |177|-----|----------|---------|178| websearch | Exa | Web search for current information |179| context7 | Context7 | Official documentation lookup |180| grep_app | Grep.app | GitHub code search |181| playwright | Browser | Browser automation |182| git-master | Git | Git operations |183184#### Custom MCPs185- Add custom MCP servers186- Configure MCP groups187- Per-project MCP settings188- MCP server authentication189190### 7. Hook System191192#### Available Hooks193194| Hook | Purpose | Usage |195|------|---------|-------|196| PreToolUse | Validate before tool execution | Security, logging |197| PostToolUse | Process after tool execution | Verification, cleanup |198| UserPromptSubmit | Process user prompts | Context injection |199| Stop | Cleanup on session end | Resource cleanup |200| PreAgentRun | Before agent execution | Setup, validation |201| PostAgentRun | After agent completion | Review, logging |202203#### Hook Configuration204- Enable/disable specific hooks205- Chain multiple handlers206- Custom hook scripts207- Per-agent hook settings208209### 8. Claude Code Compatibility210211#### Supported Features212- **Commands**: `/` commands for common actions213- **Agents**: Full agent system compatibility214- **Skills**: Skill tool integration215- **MCP**: MCP server integration216- **Hooks**: PreToolUse, PostToolUse, UserPromptSubmit, Stop217218#### Compatibility Mode219- Seamless migration from Claude Code220- Identical command interface221- Shared configuration patterns222223### 9. Session Management224225#### Session Features226- **Session List**: View all sessions227- **Session Read**: Review previous sessions228- **Session Search**: Find by content229- **Session Export**: Share or backup230231#### Session Persistence232- Configurable storage location233- Automatic session saving234- Session continuity across restarts235- Max session count management236237### 10. Todo System238239#### Features240- **TODO List**: Track all tasks241- **TODO Write**: Create/manage todos242- **Status Tracking**: Pending, in_progress, completed243- **Auto-Continuation**: Forces agent to finish244245#### Enforcement246- Strict TODO following247- Auto-reminder for incomplete tasks248- Task completion verification249- Progress reporting250251## Productivity Features252- Core operation execution with comprehensive error handling253- Input validation and output quality assurance254- Integration with existing workflows and toolchains255- Detailed logging for debugging and audit trails256257258### 1. Ralph Loop259260**Purpose**: Self-referential development loop261262**Features**:263- Automatic plan execution and validation264- Iterative improvement265- Goal tracking266- Progress monitoring267268### 2. Ultrawork Mode269270**Activation**: Include "ultrawork" or "ulw" in prompt271272**Effects**:273- Parallel agents activate274- Background tasks enabled275- Deep exploration mode276- Relentless execution277278### 3. Comment Checker279280**Purpose**: Maintain code quality281282**Rules**:283- Prevents excessive comments284- Justifies necessary comments285- Enforces code readability286- Follows project style287288### 4. Context Injection289290**Features**:291- Auto-inject AGENTS.md292- Conditional rules injection293- Project context awareness294- User preferences preservation295296### 5. Auto-Completion297298**Features**:299- End-to-end task completion300- Verification before finishing301- Evidence-based reporting302- Quality gate enforcement303304## Workflow Enhancements3051. Receive input and validate format3062. Route to appropriate handler based on input type3073. Execute core operation with monitoring3084. Transform output to expected format3095. Return results or trigger follow-up actions310311312### 1. Parallel Agent Dispatching313314```typescript315// Fire multiple agents for independent tasks316- Explore: Map codebase structure317- Librarian: Find relevant documentation318- Hephaestus: Research best practices319- Oracle: Review architecture320```321322### 2. Subagent-Driven Development323324```typescript325// Dispatch fresh subagent per task326- Two-stage review (spec, quality)327- No context pollution328- Parallel-safe execution329- Review checkpoints330```331332### 3. Git Worktree Integration333334```typescript335// Isolated workspace per task336- Create worktree for feature337- Commit to feature branch338- Merge when complete339- Clean up when done340```341342### 4. Test-Driven Development343344```typescript345// Subagents follow TDD346- Write test first347- Implement feature348- Verify tests pass349- Refactor as needed350```351352### 5. Systematic Debugging353354```typescript355// Structured debugging approach356- Reproduce issue357- Identify root cause358- Implement fix359- Verify solution360- Document learnings361```362363## Tool Categories364| Tool | Purpose | Required |365|------|---------|----------|366| CLI | Primary execution | Yes |367| API client | External service calls | Conditional |368| Validator | Output checking | Recommended |369370371### File Operations372- **read**: Read file contents373- **write**: Write/overwrite files374- **edit**: Modify file contents375- **glob**: Find files by pattern376- **grep**: Search file contents377378### Code Analysis379- **lsp_diagnostics**: Get diagnostics380- **lsp_goto_definition**: Jump to definition381- **lsp_find_references**: Find references382- **lsp_rename**: Rename symbol383- **ast_grep_search**: AST search384- **ast_grep_replace**: AST replace385386### System Operations387- **bash**: Execute shell commands388- **interactive_bash**: Interactive terminal389- **session_***: Session management390- **Notion_***: Notion integration391392### Browser Operations393- **skill_mcp**: MCP server tools394- **playwright**: Browser automation395- **dev-browser**: Persistent browser state396- **agent-browser**: CLI browser automation397- **browser-use**: Web automation398399### Web Operations400- **websearch_web_search_exa**: Web search401- **google_search**: Google search + analysis402- **webfetch**: Fetch URL content403- **look_at**: Analyze media files404405## Configuration Categories406```json407{408 "enabled": true,409 "autoRun": false,410 "timeout": 30000,411 "retries": 3412}413```414415Set environment variables as needed for authentication and endpoints.416417418### Agent Configuration419- Model selection per agent420- Temperature tuning421- Permission settings422- Custom prompts423424### Feature Configuration425- Parallel execution limits426- Background task settings427- Hook enable/disable428- Quality gate settings429430### Integration Configuration431- MCP server setup432- Hook script configuration433- Plugin integration434- External tool connection435436### UI Configuration437- Theme selection438- Keybind customization439- Layout preferences440- Display options441442## Security Features443- Core operation execution with comprehensive error handling444- Input validation and output quality assurance445- Integration with existing workflows and toolchains446- Detailed logging for debugging and audit trails447448449### Permission System450- File system access control451- Network access restrictions452- Shell command allowlisting453- API key management454455### Validation456- Input validation before execution457- Content hash verification458- Safe mode options459- Audit logging460461### Compliance462- ToS-aware configuration463- Security best practices464- Audit trail maintenance465- Secret management466467## Performance Features468- Core operation execution with comprehensive error handling469- Input validation and output quality assurance470- Integration with existing workflows and toolchains471- Detailed logging for debugging and audit trails472473474### Optimization475- Intelligent context truncation476- Token usage tracking477- Parallel execution efficiency478- Cache management479480### Scalability481- Concurrent agent limits482- Resource usage monitoring483- Memory optimization484- Background task queuing485486## Extensibility487This section covers extensibility for the oh-my-opencode-features skill.488Key operations include input validation, core processing, and output verification.489Refer to the skill overview for detailed usage instructions.490491492### Custom Tools493- Create custom tools494- Define tool schemas495- Implement execution logic496- Register with system497498### Custom Hooks499- Write custom hooks500- Script integration501- Event-driven automation502- Workflow customization503504### Plugin System505- Extend functionality506- Add new agents507- Integrate external services508- Customize behavior509510## Comparison with Other Harnesses511512| Feature | Oh My OpenCode | Standard OpenCode | Claude Code |513|---------|---------------|-------------------|-------------|514| Agents | 5 curated | Basic | 1 |515| Parallel Tasks | Yes | Limited | No |516| Background Exec | Yes | No | No |517| Hook System | 25+ | Basic | 4 |518| MCP Built-ins | 5+ | 0 | 0 |519| Claude Compat | Full | Full | Native |520| Hash Anchoring | Yes | No | No |521| TODO Enforcement | Yes | No | No |522523## Feature Quick Reference524525| Category | Key Features |526|----------|-------------|527| Agents | Sisyphus, Hephaestus, Oracle, Librarian, Explore |528| Models | OpenAI, Anthropic, Google, OpenCode Zen |529| Tools | LSP, AST-Grep, File Ops, Bash, Session |530| MCPs | websearch, context7, grep_app, playwright, git-master |531| Hooks | 25+ hooks, custom scripts, per-agent settings |532| Sessions | List, Read, Search, Export, Persistence |533| Workflow | Ultrawork, Ralph Loop, TDD, Debugging |534| Quality | Comment Checker, TODO Enforcement, Verification |535536## The Process5375381. **Identify feature needs** – Determine which oh-my-opencode features address your task (agents, MCPs, hooks, sessions, workflows)5392. **Verify feature availability** – Check feature is enabled in configuration and compatible with your OpenCode version5403. **Configure feature** – Set up MCP servers, hooks, categories as needed in config5414. **Load feature in session** – Ensure feature is accessible (agents via /agent, MCPs via /mcp, etc.)5425. **Use feature** – Execute feature-specific workflows (parallel agents, hooks, ultrawork, etc.)5436. **Monitor execution** – Watch feature behavior, adjust configs, debug issues5447. **Verify output** – Run lsp_diagnostics, build, tests to confirm feature completed correctly545546## When NOT to Use547548- Task is outside your authorization scope549- You need to implement controls (use implementing-* skills)550- Task is about analysis, not action (use analyzing-* skills)551- You don't have access to target systems552- Task requires compliance expertise (consult professionals)553- Task is about defense, not offense (use defensive skills)554555556## Red Flags557558- **Feature conflicts** – Enabling multiple conflicting features (e.g., multiple LSP tools fighting over same file)559- **Version mismatch** – Using features from different oh-my-opencode versions causing incompatibilities560- **Missing dependencies** – Enabling MCPs without required tools or permissions561- **Hook loops** – Hooks triggering themselves infinitely due to improper conditions562- **Over-enforcement** – Enabling all quality checks when only specific ones needed causes slowdowns563- **Ignoring workflows** – Not following Ultrawork, Ralph Loop, or TDD when they should apply564- **Session corruption** – Attempting to use features on stale or corrupted sessions565566## Verification567568- **Feature loads**: Feature commands work (e.g., `/mcp list`, `/agent sisyphus`)569- **MCPs respond**: MCP servers execute tool calls and return expected results570- **Hooks fire**: Hook conditions trigger expected actions at right times571- **Sessions valid**: Session commands return valid data without errors572- **Workflows execute**: Ultrawork completes autonomously, TDD creates tests, Ralph Loop iterates to completion573- **Quality checks pass**: Comment checker, TODO enforcement, verification run as configured574- **Integration clean**: All features work together without conflicts or race conditions575576## References577578- Features Overview: https://github.com/code-yeongyu/oh-my-opencode#features579- Configuration: https://github.com/code-yeongyu/oh-my-opencode/blob/master/docs/configurations.md580- Agent System: https://github.com/code-yeongyu/oh-my-opencode#for-those-who-want-to-read-meet-sisyphus581- OpenCode Docs: https://opencode.ai/docs/582583## Anti-Rationalization Table584585| Rationalization | Reality |586|---|---|587| "I will handle auth later" | Retrofitting auth is 10x harder. Build it from day one. |588| "APIs do not change" | APIs change. Version your integrations and handle deprecations. |589| "Webhooks are optional" | Without webhooks, you miss real-time events. They are essential. |590591## Related Skills592593- oh-my-opencode: Overall integration594- oh-my-opencode-installation: Setup guide595- oh-my-opencode-agents: Agent deep dive596- oh-my-opencode-usage: Daily usage597- oh-my-opencode-configuration: Advanced setup598599## Process6006011. Analyze the task requirements6022. Apply domain expertise6033. Verify output quality