### Initial Setup and Initialization Commands

Source: https://context7.com/oimiragieo/agent-studio/llms.txt

Run these commands to install dependencies, compile registries, initialize memory, reindex code, and validate the setup. Remember to copy and populate the .env file with your API keys.

```bash
# Initial setup (installs deps, compiles registries, indexes code)
pnpm run setup

# Initialize memory database (SQLite schema creation)
pnpm run memory:init

# Regenerate agent registry after adding/modifying agents
pnpm run agents:registry

# Build full hybrid search index (BM25 + semantic vector)
# ~12 min with GPU, ~17 min CPU-only
pnpm run code:index:reindex

# Verify everything is wired correctly
pnpm run validate:full

# Copy environment config and add API keys
cp .env.example .env
```

--------------------------------

### Copy Example Environment File

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Copy the example environment file to start configuring your API keys. This is the first step for setting up optional API keys for AI-powered skills.

```bash
cp .env.example .env
```

--------------------------------

### Drop-In Setup for Existing Projects

Source: https://context7.com/oimiragieo/agent-studio/llms.txt

Integrate Agent Studio into an existing project by copying the .claude directory, installing dependencies, and initializing core artifacts. Set the ANTHROPIC_API_KEY for memory operations.

```bash
# 1. Copy .claude/ into the target repository root
cp -r /path/to/agent-studio/.claude /path/to/your-project/
```

```bash
# 2. Install Node.js dependencies (from your project root)
pnpm install
```

```bash
# 3. Initialize core artifacts
pnpm run memory:init
pnpm run agents:registry
pnpm run routing:prototypes
```

```bash
# 4. Optional: build code search index
pnpm run code:index:reindex
```

```bash
# 5. Set required environment variable
cp .env.example .env
# Edit .env: set ANTHROPIC_API_KEY for memory extraction/dedup
```

--------------------------------

### Validate Onboarding Setup

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Run the onboarding tool to validate your setup. This is recommended for first-time users.

```bash
node .claude/tools/onboarding.mjs
```

--------------------------------

### CLI Tool Test Coverage Example

Source: https://github.com/oimiragieo/agent-studio/blob/main/tests/CLAUDE.md

Outlines the tests for CLI tools, specifying which files cover CLI output formatting and package manager detection within the setup wizard.

```markdown
| Test File                        | What It Covers                             |
| -------------------------------- | ------------------------------------------ |
| `memory-dashboard.test.cjs`      | Memory dashboard CLI output and formatting |
| `setup-package-manager.test.cjs` | Package manager detection in setup wizard  |
```

--------------------------------

### Run First-Time Setup

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/CLAUDE.md

Execute this command once to set up the project environment.

```bash
pnpm run setup
```

--------------------------------

### Get Help for Rule Auditor Command

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Use the --help flag with any slash command to get detailed usage information. This example shows how to get help for the rule-auditor command.

```shell
/rule-auditor --help
```

--------------------------------

### Install Gemini CLI

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Install the Gemini CLI globally and verify the installation. This is a required step for using Codex skills. Authentication can be done via API key or session.

```bash
# Install globally
npm install -g @anthropic-ai/gemini
```

```bash
# Verify installation
gemini --version
```

```bash
# Authenticate (requires Google Cloud API key)
export GEMINI_API_KEY=your-api-key-here
# or use session-based auth
gemini login
```

--------------------------------

### Install Latest Version and Dependencies

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Pull the latest code changes and install project dependencies.

```bash
git pull && pnpm install
```

--------------------------------

### Configure and Start Telegram Integration

Source: https://context7.com/oimiragieo/agent-studio/llms.txt

Set up environment variables for Telegram bot token and owner ID, then start the Telegram daemon. Alternatively, use in-session commands to manage the integration.

```bash
# Configure .env:
TELEGRAM_BOT_TOKEN=<token from @BotFather>
TELEGRAM_OWNER_ID=<your numeric ID from @userinfobot>
TELEGRAM_ALLOWED_USERS=<your numeric ID>
CHANNEL_AUTO_START=true

# Start the Telegram daemon
pnpm run start:telegram

# Or from within a Claude Code session:
/enable-telegram
/disable-telegram
/restart-telegram

# Key bot commands (send from Telegram):
# /status    — active loops, pending tasks, last heartbeat
# /tasks     — current task list with status
# /memory QUERY  — search agent memory for a keyword
# /ask QUESTION  — ask the AI and get a reply
# /spawn TYPE DESC  — spawn an agent (general-assistant, researcher, etc.)
# /code TASK — mission-aware coding (classifies task → 16 agent types → TDD → grade 0-100)

# HTTP API (runs at localhost:3101 when daemon is active):
curl http://127.0.0.1:3101/status
curl http://127.0.0.1:3101/history
curl -X POST http://127.0.0.1:3101/send -d '{"message":"run tests"}'

# Webhook endpoint for GitHub/CI events:
curl -X POST http://127.0.0.1:3101/webhook \
  -H "Content-Type: application/json" \
  -d '{"event":"push","repo":"myrepo","branch":"main"}'
```

--------------------------------

### Install Dependencies and Initialize Memory

Source: https://github.com/oimiragieo/agent-studio/blob/main/GEMINI.md

Run these commands to set up the project, initialize the memory database, and prepare agent registries and routing.

```bash
pnpm install
pnpm memory:init
pnpm agents:registry
pnpm routing:prototypes
pnpm agents:catalog
```

--------------------------------

### Install GitHub Copilot CLI (Optional)

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Install the GitHub Copilot CLI globally and authenticate with GitHub. This is an optional step for using Codex skills.

```bash
# Install globally
npm install -g @github/copilot
```

```bash
# Authenticate with GitHub
gh auth login
```

```bash
# Verify installation
copilot --version
```

--------------------------------

### Install MarkItDown for File Conversion

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Install the MarkItDown library with all extras using pip to enable file conversion to Markdown for agent memory.

```bash
pip install 'markitdown[all]'
```

--------------------------------

### Verify Search Tool Installation

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Confirms that rga, fzf, and sg (ast-grep shim) have been installed correctly by checking their version numbers.

```powershell
rga --version
fzf --version
sg --version
```

--------------------------------

### Install OpenAI Codex CLI (Optional)

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Install the OpenAI Codex CLI globally and verify the installation. This is an optional step for using Codex skills.

```bash
# Install globally
npm install -g @openai/codex
```

```bash
# Verify installation
codex --version
```

```bash
# Authenticate
export OPENAI_API_KEY=your-api-key-here
```

--------------------------------

### Start, Status, and Stop Search Daemon

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Commands to manage the search daemon. Use these to start, check the status of, or stop the background search process.

```bash
pnpm search:daemon:start
```

```bash
pnpm search:daemon:status
```

```bash
pnpm search:daemon:stop
```

--------------------------------

### Prewarm Search Daemon and Indexes

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Prepares the search daemon, LanceDB, and semantic path for faster initial queries. Run this after starting the daemon.

```bash
pnpm search:daemon:prewarm
```

--------------------------------

### Install Verified Skills from Marketplace

Source: https://context7.com/oimiragieo/agent-studio/llms.txt

Install verified skill packages from the marketplace using the provided command. This process includes security checks like HMAC-SHA256 signature verification and trust score assessment.

```bash
# Install a verified skill package from the marketplace
pnpm run skill:install <package-name>
# Verifies HMAC-SHA256 signature, checks 4-tier trust score,
# path-traversal guards block ../ and URL-encoded variants
```

--------------------------------

### Install and Run Validation Scripts

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Install project dependencies and run validation scripts to check configuration, models, workflows, and references. Use `pnpm validate` for fast checks or `pnpm validate:all` for full validation.

```bash
# Install validation dependencies
pnpm install

# Run validation
pnpm validate          # Fast validation (config, models)
pnpm validate:all      # Full validation (includes workflows, references, CUJs, rule index)

# Cleanup (recommended before commits)
pnpm cleanup:check     # Preview what would be deleted (safe)
pnpm cleanup           # Delete temp/runtime files (use with care)
node .claude/tools/cleanup-repo.mjs --dry-run --reports-retention-days 5   # Optional: preview pruning reports older than 5 days
node .claude/tools/cleanup-repo.mjs --execute --reports-retention-days 5   # Optional: prune reports older than 5 days (opt-in)
```

--------------------------------

### Install Scoop for Windows

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Installs Scoop, a command-line installer for Windows, for non-administrator PowerShell environments. Ensure your execution policy is set to RemoteSigned.

```powershell
Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser
Invoke-RestMethod -Uri https://get.scoop.sh | Invoke-Expression
```

--------------------------------

### Initialize Agent Studio Core Artifacts

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Execute these pnpm commands to initialize the core artifacts when dropping Agent Studio into another repository. Ensure Node.js and pnpm are installed.

```bash
pnpm memory:init
pnpm agents:registry
pnpm routing:prototypes
```

--------------------------------

### Cross-Platform Handoff Examples

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Illustrates how to seamlessly transfer context and tasks between different development environments like Claude, Cursor, and Factory Droid. This enables a fluid workflow across multiple tools.

```text
# Claude to Cursor
"Sync this context to Cursor for implementation"
```

```text
# Cursor to Factory
Use @context-bridge to hand off to Factory
```

```text
# Factory to Claude
Run Task tool with skill context-bridge to sync to Claude
```

--------------------------------

### Automatic Keyword Matching Example

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Illustrates how user prompts are processed to detect keywords, match them against configured workflows, and select the highest priority matching workflow.

```text
User: "Build a new authentication feature from scratch"
→ Keywords detected: "new", "from scratch"
→ Matches: `fullstack` workflow (keywords: "new project", "from scratch", "greenfield")
→ Priority: 2 (medium priority)
→ Selected: `fullstack` workflow
→ Executes: Planner (Step 0) → Full Stack Flow with all agents (Steps 1-N)
```

--------------------------------

### Install Cursor Agent (Optional)

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Install the Cursor Agent via WSL on Windows. This is an optional step for using Codex skills.

```bash
# Inside WSL
wsl bash -lc "curl https://cursor.com/install -fsS | bash"
```

```bash
# Verify installation
cursor-agent --version
```

--------------------------------

### Install Claude Code CLI

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Install the Claude Code CLI globally and verify the installation. This is a required step for using Codex skills.

```bash
# Install globally
npm install -g @anthropic-ai/claude-code
```

```bash
# Verify installation
claude --version
```

```bash
# Authenticate
claude login
```

--------------------------------

### Install Skill Package via MMP CLI

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Command to install a skill package using the Memory Marketplace Platform CLI. This command verifies the package signature and trust score before installation.

```bash
pnpm skill:install <package>
```

--------------------------------

### Initiate Code Quality Review

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Start a comprehensive code review and refactoring process using the /code-quality command.

```bash
/code-quality
```

--------------------------------

### Start Quality Daemon Continuously

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Run the Autonomous Quality Daemon in a continuous loop in the foreground. This ensures ongoing quality checks and monitoring.

```bash
pnpm quality:daemon:start
```

--------------------------------

### Install Search Tools via Scoop

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Installs essential search utilities: rga (ripgrep-all) for searching archives and various file types, fzf for interactive fuzzy finding, and ast-grep for structural code analysis.

```powershell
# Install rga (ripgrep-all)
scoop install rga

# Install fzf
scoop install fzf

# Install ast-grep (includes `sg` shim)
scoop install ast-grep
```

--------------------------------

### Configuration Test Coverage Example

Source: https://github.com/oimiragieo/agent-studio/blob/main/tests/CLAUDE.md

Specifies the test file for configuration validation, focusing on ensuring consistency in phase model assignments.

```markdown
| Test File                           | What It Covers                                   |
| ----------------------------------- | ------------------------------------------------ |
| `phase-models-consistency.test.cjs` | Validates phase model assignments are consistent |
```

--------------------------------

### Configure Daemon Auto-Prewarm

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Enables automatic prewarming when the search daemon starts. Set `HYBRID_DAEMON_PREWARM` to `true`.

```bash
HYBRID_DAEMON_PREWARM=true pnpm search:daemon:start
```

--------------------------------

### Mobile Development Prompt

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Start a mobile development project with this prompt. The process includes platform strategy, UI design, feature implementation, and cross-device testing.

```bash
"/mobile"
```

--------------------------------

### Test File Structure Example

Source: https://github.com/oimiragieo/agent-studio/blob/main/tests/CLAUDE.md

Illustrates the hierarchical structure of the test suite, mirroring the project's source directories. This organization helps in locating tests for specific modules.

```tree
tests/
├── agents/             # Agent definition tests
├── cli/                # CLI tool tests
├── code-indexing/      # Code search and indexing tests
├── config/             # Configuration validation tests
├── hooks/              # Hook behavior tests
├── integration/        # Cross-component integration tests
├── lib/                # Library module tests (mirrors .claude/lib/)
│   ├── agents/         # Agent config schema tests
│   ├── code-indexing/  # BM25, embedding, hybrid search internals
│   ├── creators/       # Creator ecosystem impact tests
│   ├── memory/         # Memory system tests (LanceDB, tiers, pruning)
│   ├── monitoring/     # Metrics, health, SLO monitoring tests
│   ├── party-mode/     # Consensus voting tests
│   ├── plan/           # Plan progress tracking tests
│   ├── qa/             # QA criteria and report tests
│   ├── reflection/     # Reflection system contract tests
│   ├── routing/        # Routing logic tests
│   ├── self-healing/   # Loop state management tests
│   ├── spawn/          # Spawn template and assembly tests
│   ├── text-processing/# Sentence chunker tests
│   ├── tools/          # Tool library tests
│   ├── utils/          # Utility module tests
│   └── workflow/       # Workflow engine and state tests
├── misc/               # Miscellaneous validation tests
├── performance-profiling-minimal.test.cjs
├── phase-4/            # Phase 4 workflow pattern tests
├── routing-table.test.cjs
├── scale/              # Scale and performance tests
├── scripts/            # Script import and security tests
├── skills/             # Skill creator and lifecycle tests
├── spec-init.test.cjs
├── tools/              # Tool integration tests
├── unit/               # UI formatter and unit tests
├── utils/              # Token budget and utility tests
├── workflows/          # Workflow creator/updater tests
├── fixtures/           # Static test data (JSON, .gitkeep)
└── (various root-level .test.cjs files)

```

--------------------------------

### Cross-Platform Skill Invocation Examples

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Demonstrates how to invoke agent skills across different platforms like Claude, Cursor IDE, and Factory Droid. Each platform uses a distinct syntax for accessing functionalities such as rule selection, scaffolding, and auditing.

```text
# Natural language invocation
"Select rules for this project"
"Scaffold a UserProfile component"
"Audit src/components/ against our rules"
```

```text
# @ symbol invocation
Use @rule-selector to configure rules
Use @scaffolder to create a new component
Use @rule-auditor to check this code
```

```text
# Task tool invocation
Run Task tool with skill rule-selector
Run Task tool with skill scaffolder
Run Task tool with skill rule-auditor
```

--------------------------------

### Build New Feature Prompt

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Use this prompt to initiate the workflow for building a new feature, such as a task management dashboard with user authentication. The system will select the 'fullstack' workflow and guide through planning, design, development, and QA.

```bash
"I need a task management dashboard with user authentication"
```

--------------------------------

### Agent Test Coverage Example

Source: https://github.com/oimiragieo/agent-studio/blob/main/tests/CLAUDE.md

Details the test files within the `tests/agents/` directory and the specific aspects of agent behavior they cover. This includes planner agent functionality.

```markdown
| Test File               | What It Covers                                      |
| ----------------------- | --------------------------------------------------- |
| `core/planner.test.cjs` | Planner agent task creation and dependency handling |
| `*.archived`            | Archived agent tests (not run)                      |
```

--------------------------------

### Register Hooks in settings.json

Source: https://context7.com/oimiragieo/agent-studio/llms.txt

Example configuration for registering various hooks like UserPromptSubmit, PreToolUse, PostToolUse, and SessionEnd in .claude/settings.json. Hooks can be filtered by 'matcher' strings.

```json
// settings.json hook registration (abbreviated)
{
  "hooks": {
    "UserPromptSubmit": [
      {
        "matcher": "",
        "hooks": [
          { "type": "command", "command": "node .claude/hooks/reflection/force-step0-execution.cjs" },
          { "type": "command", "command": "node .claude/hooks/session/user-prompt-orchestrator.cjs" }
        ]
      }
    ],
    "PreToolUse": [
      {
        "matcher": "",
        "hooks": [
          { "type": "command", "command": "node .claude/hooks/routing/pre-tool-unified.cjs" }
        ]
      },
      {
        "matcher": "Edit|Write|NotebookEdit",
        "hooks": [
          { "type": "command", "command": "node .claude/hooks/routing/unified-creator-guard.cjs" },
          { "type": "command", "command": "node .claude/hooks/safety/write-pretool-bundle.cjs" }
        ]
      },
      {
        "matcher": "Bash",
        "hooks": [
          { "type": "command", "command": "node .claude/hooks/safety/bash-pretool-bundle.cjs" }
        ]
      }
    ],
    "PostToolUse": [
      {
        "matcher": "",
        "hooks": [
          { "type": "command", "command": "node .claude/hooks/metrics/post-tool-metrics-unified.cjs" }
        ]
      },
      {
        "matcher": "Edit|Write|NotebookEdit",
        "hooks": [
          { "type": "command", "command": "node .claude/hooks/memory/sync-memory-index.cjs" }
        ]
      }
    ],
    "SessionEnd": [
      {
        "matcher": "",
        "hooks": [
          { "type": "command", "command": "node .claude/hooks/reflection/unified-reflection-handler.cjs" }
        ]
      }
    ]
  }
}
```

--------------------------------

### Migrate Agent Studio from v2.x to v3.0

Source: https://context7.com/oimiragieo/agent-studio/llms.txt

Follow these steps to migrate your Agent Studio project to v3.0. This includes installing dependencies, previewing changes, applying the migration, and regenerating the agent registry.

```bash
# 1. Pull latest and install
git pull && pnpm install
```

```bash
# 2. Preview migration changes (dry run — no files written)
pnpm run migrate:2x-to-3 --dry-run
```

```bash
# 3. Apply migration (backfills agent manifests, flags SSE transport)
pnpm run migrate:2x-to-3
```

```bash
# 4. Review backups created for modified agents
ls .claude/context/tmp/agents-pre-v3-migration/
```

```bash
# 5. Update any mcp.transport: "sse" → "streamable-http" entries in config
#    (script flags locations but does not rewrite config files)
```

```bash
# 6. Regenerate agent registry in v3 schema format
pnpm run agents:registry
```

```bash
# 7. Enable enforcement when ready (optional — off by default)
#    V3_MANIFEST_REQUIRED=on   (in .env)
```

```bash
# 8. Verify
pnpm run test:framework
```

--------------------------------

### Code Indexing Test Coverage Example

Source: https://github.com/oimiragieo/agent-studio/blob/main/tests/CLAUDE.md

Lists the test files for the code indexing system, detailing coverage for AST-grep integration, hybrid search CLI, and core search algorithms.

```markdown
| Test File                      | What It Covers                               | 
| ------------------------------ | -------------------------------------------- |
| `ast-grep-wrapper.test.cjs`    | AST-grep wrapper for structural code search  |
| `cli.test.cjs`                 | Hybrid search CLI interface                  |
| `hybrid-search-cli.test.cjs`   | End-to-end CLI hybrid search behavior        |
| `hybrid-search.test.cjs`       | Core hybrid search (BM25 + semantic ranking) |
| `parser.test.cjs`              | Code parser for chunk extraction             |
| `query-analyzer.test.cjs`      | Query intent classification                  |
| `result-ranker.test.cjs`       | RRF result merging and ranking               |
| `ripgrep-integration.test.cjs` | ripgrep integration for text search          |
| `semantic-chunker.test.cjs`    | Semantic chunking of source files            |
```

--------------------------------

### Advanced Interactive ripgrep Launcher

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Launches fzf with ripgrep integration, allowing dynamic reloading of search results as the user types. The `rg_prefix` variable sets up the ripgrep command, and `--bind` configures fzf's behavior for starting and changing search queries.

```bash
: | rg_prefix='rg --column --line-number --no-heading --color=always --smart-case' \
  fzf --ansi --disabled \
      --bind 'start:reload:$rg_prefix ""' \
      --bind 'change:reload:$rg_prefix {q} || true'
```

--------------------------------

### Manage Agent Costs with Cost Tracker

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Use the cost-tracker tool to monitor token usage and estimate costs for agent sessions. This includes viewing reports, getting stats for specific agents, and forecasting future expenses.

```bash
# View cost report
node .claude/tools/cost-tracker.mjs report

# Get cost stats for specific agent
node .claude/tools/cost-tracker.mjs stats developer 30

# Get cost forecast
node .claude/tools/cost-tracker.mjs forecast 30
```

--------------------------------

### Display Help Instructions

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/validation/README-VERSION-VALIDATION.md

Run this command to display the usage instructions and available options for the validation script.

```bash
node scripts/validate-rule-index-paths.mjs --help
```

--------------------------------

### Configuration Loading Details

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/daemon/CLAUDE.md

Explains how the configuration is loaded from environment variables and JSON files, including user access control.

```javascript
Reads from `.env` (via `loadDotenv`) and `~/.claude/channels/config.json`.
```

```javascript
Builds allowed users set from `TELEGRAM_ALLOWED_USERS`, `TELEGRAM_OWNER_ID`, and `~/.claude/channels/telegram/access.json`.
```

--------------------------------

### Start A2A Discovery Server

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Start the A2A discovery server to serve AgentCards and enable external agent discovery. The server listens for AgentCards at /.well-known/agent-card.json.

```plaintext
"Start A2A discovery server"
```

--------------------------------

### Analyze Token Budget and Refactor Recommendations

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Use this command for token budget analysis within a specified directory. It provides file/directory sizes, token estimates, and recommends splitting oversized source files (>15K tokens) for better AI agent readability.

```bash
pnpm search:tokens .claude/lib
```

--------------------------------

### Adding New Platforms - Daemon Implementation

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/README.md

Steps to add new platforms like Discord or Slack by implementing specific methods and configuration files within the daemon structure.

```javascript
1. Create `daemon/sources/discord.cjs` implementing `start()` and `stop()` methods
2. Create `daemon/sinks/discord.cjs` implementing `send(chatId, text, opts)`
3. Add config section in `daemon/config.cjs`
4. Wire in `daemon/index.cjs`
5. The router, dispatcher, memory, and renderer are platform-agnostic
```

--------------------------------

### Copy Agent Studio Configuration (Bash)

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Clone or download the Agent Studio repository and copy the configuration files to your project directory using bash commands.

```bash
cp -r .claude/ /path/to/your/project/
cp .claude/CLAUDE.md /path/to/your/project/CLAUDE.md
```

--------------------------------

### Worker Summary

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Run this command to get a summary of the worker's last ticks and status.

```bash
pnpm worker:summary
```

--------------------------------

### Run All Tests

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/PLAN-PHASE5.md

Execute the full test suite for the project. This is a comprehensive check performed after Phase 5 is complete.

```bash
pnpm test
```

--------------------------------

### Memory System Operations

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Commands related to memory lineage and the Skill Marketplace. These are used for managing memory records and installing skills.

```bash
pnpm mmp:lineage <record-id>      # walk ancestry chain for a CAT7 memory record
```

```bash
pnpm mmp:descendants <record-id>  # list all downstream records
```

```bash
pnpm skill:install <package>      # install a verified skill package from the marketplace
```

--------------------------------

### Analyze Project Structure and Dependencies

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

This command displays the project structure, dependencies, and generates a Mermaid diagram for visualization. It helps in understanding the project's architecture.

```bash
pnpm search:structure
```

--------------------------------

### Trigger Memory Consolidation via HTTP API

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/README.md

Initiate the memory consolidation process by making a GET request to the `/dream` endpoint.

```bash
curl http://127.0.0.1:3101/dream
```

--------------------------------

### Artifact Naming with Template Variables

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Provides an example of how to use template variables, such as `{{workflow_id}}`, in artifact filenames to ensure unique and context-aware naming.

```yaml
outputs:
  - plan-{{workflow_id}}.json
  # Resolves to: plan-1735123456-abc123.json
```

--------------------------------

### Control Telegram Daemon via CLI

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/README.md

Use these commands to start, stop, check the status, or restart the Telegram channel daemon from the command line.

```bash
node scripts/channels/telegram-ctl.cjs start
node scripts/channels/telegram-ctl.cjs status
node scripts/channels/telegram-ctl.cjs stop
node scripts/channels/telegram-ctl.cjs restart
```

--------------------------------

### Configure Context Budget Thresholds

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Set the warning threshold in tokens for context size before a session reaches the compression threshold. Enable hard-blocking for spawns that exceed 1.6x the threshold.

```bash
SPAWN_BUDGET_DEFAULT_CONTEXT=50000
```

```bash
SPAWN_BUDGET_HARD=on
```

--------------------------------

### Telegram Control Script Commands

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/daemon/CLAUDE.md

Demonstrates the command-line interface for controlling the Telegram bot, including starting, stopping, checking status, and performing diagnostic checks.

```bash
start, stop, status, restart, doctor (validate config), doctor --fix (validate + auto-repair)
```

--------------------------------

### Run Comprehensive Test Suite

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/validation/README-VERSION-VALIDATION.md

Execute this command to run the full test suite for version validation, ensuring all scenarios are covered.

```bash
pnpm test:version-validation
```

--------------------------------

### Memory Marketplace CLI

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Command-line interface for interacting with the Memory Marketplace.

```APIDOC
## Memory Marketplace CLI

### Description
Tools for inspecting and managing memory records and skill packages.

### Commands
- `pnpm mmp:lineage <record_id>`: Inspect the derivation graph of a memory record.
- `pnpm mmp:descendants <record_id>`: Inspect downstream records influenced by a memory record.
- `pnpm skill:install <package>`: Install a verified skill package after signature and trust score verification.
```

--------------------------------

### Control Telegram Daemon via Claude Code Skills

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/README.md

Manage the Telegram daemon using these commands within Claude Code skills for setup, enabling, and disabling.

```bash
/setup-telegram     # verify config
/enable-telegram    # start daemon
/disable-telegram   # stop daemon
```

--------------------------------

### Initialize and Manage Memory System

Source: https://context7.com/oimiragieo/agent-studio/llms.txt

Commands to initialize the SQLite schema, check memory health, view the memory dashboard, run weekly maintenance, and backfill legacy JSON files into the entity index.

```bash
# Initialize SQLite schema (run once)
pnpm run memory:init
```

```bash
# Check memory health
pnpm run memory:health
# Output: { status: "ok", stmCount: 12, mtmCount: 45, ltmCount: 8, ... }
```

```bash
# View memory dashboard (token budgets, SLO metrics)
pnpm run memory:dashboard
```

```bash
# Run weekly maintenance (compaction, archival)
pnpm run memory:weekly
```

```bash
# Backfill patterns.json / gotchas.json into SQLite entity index
pnpm run memory:sync-json
```

```bash
# Archive oversized learnings.md (triggers at 40KB)
node .claude/lib/memory/memory-manager.cjs archive-learnings
```

--------------------------------

### Start Heartbeat Loops Manually

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Initiate heartbeat loops manually within your Claude Code session using this command. This ensures the agent runtime remains healthy and indexed.

```bash
/heartbeat-start
```

--------------------------------

### Preview Migration Changes

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Run the migration script in dry-run mode to preview changes without writing any files.

```bash
pnpm migrate:2x-to-3 --dry-run
```

--------------------------------

### Wire Tick Engine into Daemon Index

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/PLAN-PHASE5.md

Conditionally initializes and starts the TimerSource (acting as a tick engine) if proactive mode is enabled in the configuration. It enqueues events via the dispatcher.

```javascript
if (config.proactive?.enabled) {
  const tickEngine = new TimerSource(config.proactive, event => dispatcher.enqueue(event));
  sources.push(tickEngine);
  tickEngine.start();
  log('Proactive tick engine started');
}
```

--------------------------------

### Skill Marketplace Environment Variables

Source: https://context7.com/oimiragieo/agent-studio/llms.txt

Configure the skill marketplace by setting these environment variables. Ensure SKILL_MARKETPLACE_HMAC_KEY is at least 32 characters long.

```bash
SKILL_MARKETPLACE_HMAC_KEY=<min 32 chars>
```

```bash
SKILL_MARKETPLACE_MIN_TRUST=0.7          # 0.0–1.0 trust threshold
```

```bash
SKILL_MARKETPLACE_REGISTRY_URL=https://...
```

--------------------------------

### Non-Blocking Task Execution Example

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/daemon/CLAUDE.md

Illustrates the asynchronous nature of task execution, where tasks spawned by Claude are handled in a background TaskPool, allowing immediate continuation of message processing.

```javascript
claudeAsync() in `claude-cli.cjs` — Non-blocking `spawn`-based wrapper alongside `claudeSync()`. Returns `{ child, promise, cancel }`.
```

```javascript
executeTaskAsync() / executeRalphLoopAsync() in `executor.cjs` — Async task execution methods. Rate limit retry uses `setTimeout` (not busy-wait).
```

```javascript
Pool event handlers in `dispatcher._wirePoolEvents()` — Deliver results, update memory, extract skills when background tasks complete.
```

--------------------------------

### Interactive File and Line Picker with Preview

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Uses rg to find lines containing 'auth|token|session' and pipes the output to fzf for interactive selection. The preview window uses 'bat' to display the content of the selected file with syntax highlighting and line numbers.

```powershell
rg --line-number --no-heading --color=always "auth|token|session" .
  | fzf --ansi --delimiter ":" \
    --preview "bat --color=always --style=numbers --highlight-line {2} {1}"
```

--------------------------------

### Directly Trigger Workflow with Slash Command

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

When a natural language prompt doesn't trigger a workflow, use a slash command for direct invocation. This example directly triggers the code-quality workflow.

```shell
/code-quality  # Directly triggers code-quality workflow
```

--------------------------------

### Agent Manifest Frontmatter Schema

Source: https://context7.com/oimiragieo/agent-studio/llms.txt

Example YAML frontmatter for an agent file, defining its metadata, capabilities, model, context strategy, and permissions. This schema is required for agent manifest v3.0+.

```yaml
# .claude/agents/core/developer.md (frontmatter)
---
verified: true
lastVerifiedAt: 2026-02-22T19:27:49.887Z
name: developer
version: 1.1.0
description: >-
  TDD-focused implementer. Writes code, runs tests, and refactors.
  Follows Red-Green-Refactor strictly. Uses ripgrep for fast code discovery.
model: sonnet
temperature: 0.3
context_strategy: lazy_load
maxTurns: 18
permissionMode: default
isolation: worktree          # spawns in git worktree for isolation
priority: high
tools:
  - Read
  - Write
  - Edit
  - Glob
  - Grep
  - Bash
  - MemoryRecord
  - TaskUpdate
  - Skill
skills:
  - tdd
  - debugging
  - code-quality-expert
  - typescript-expert
  - react-expert
  # ... 100+ available skills
---
```

--------------------------------

### Daemon File Structure

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/README.md

Overview of the daemon's directory structure and key files, including main entry points, configuration, routing, dispatching, rendering, execution, commands, memory, skills, sources, and sinks.

```tree
daemon/
├── index.cjs        # Main daemon + HTTP server
├── config.cjs       # Config loader
├── router.cjs       # Event → route matching
├── dispatcher.cjs   # Queue + deliver + tags + rate limit + interviews
├── renderer.cjs     # Claude -p + model routing + personality + skill inject
├── executor.cjs     # Task/Ralph/Ultrawork execution + rate limit retry
├── commands.cjs     # Bot /slash commands (24 commands)
├── memory.cjs       # 3-tier KAIROS memory
├── skills.cjs       # Skill extraction engine (learns from tasks)
├── sources/
│   ├── telegram.cjs # Telegram polling
│   ├── webhook.cjs  # Webhook source (POST /webhook)
│   └── timer.cjs    # KAIROS tick/heartbeat engine
└── sinks/
    └── telegram.cjs # Telegram delivery (sendMessage + sendDocument)

telegram-relay.mjs   # MCP server (tools-only in main session)
telegram-ctl.cjs     # CLI: start/stop/status/restart
```

--------------------------------

### Create New Test Files

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/PLAN.md

This outlines the file structure for new test files to be created within the `tests/channels/daemon/` directory. These files correspond to specific slices and functionalities.

```bash
tests/channels/
├── daemon/
│   ├── router.test.cjs          # Slice 0.1
│   ├── memory.test.cjs          # Slice 0.1
│   ├── commands.test.cjs        # Slice 0.1
│   ├── dispatcher.test.cjs      # Slice 0.1
│   ├── config.test.cjs          # Slice 0.1
│   ├── renderer.test.cjs        # Slice 1.2, 1.3, 2.3
│   ├── executor.test.cjs        # Slice 1.4
│   ├── smoke.test.cjs           # Slice 0.2
│   ├── sources/
│   │   └── telegram.test.cjs    # Slice 0.1
│   └── sinks/
│       └── telegram.test.cjs    # Slice 1.1, 1.2
```

--------------------------------

### Integration Smoke Test for Channel Daemon

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/PLAN.md

Integration smoke test file for the channel daemon. This test verifies the daemon's ability to start, process events via HTTP, and report status and health.

```text
tests/channels/daemon/smoke.test.cjs

```

--------------------------------

### Manual Workflow Initialization

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Manually initialize a workflow session by setting a unique WORKFLOW_ID and creating the necessary directory structure for gate files.

```bash
WORKFLOW_ID=$(date +%s)-$(uuidgen | cut -d'-' -f1)
mkdir -p .claude/context/history/gates/$WORKFLOW_ID
```

--------------------------------

### Run Quality Daemon Once

Source: https://github.com/oimiragieo/agent-studio/blob/main/README.md

Execute a single cycle of the Autonomous Quality Daemon. This is useful for testing or performing a one-off quality check.

```bash
pnpm quality:daemon:run-once
```

--------------------------------

### Agent Studio Observability Commands

Source: https://context7.com/oimiragieo/agent-studio/llms.txt

Utilize these commands for monitoring agent behavior, token consumption, and costs. The session audit provides per-component token burn, and metrics CI gates enforce performance SLOs.

```bash
# Inspect per-component token burn for a session
pnpm run session:audit <session-id>
# Output: colored table of token consumption by agent/skill/tool

# Real-time cost status (written to .claude/context/runtime/ccusage-status.txt):
# [tokens]      57,685 today (in: 1,403 / out: 56,282) | Cost: $86.82
# [cache]       $316.97 saved | 66,701,262 reads, 7,961,389 writes
# [compression] 18 events | 596.2KB freed (~152,627 tokens) | ~$0.76 saved

# Pricing table (March 2026):
# Model       Input     Output    Cache Write  Cache Read
# Opus 4.6   $5.00/M  $25.00/M   $6.25/M      $0.50/M
# Sonnet 4.6 $3.00/M  $15.00/M   $3.75/M      $0.30/M
# Haiku 4.5  $1.00/M   $5.00/M   $1.25/M      $0.10/M

# Spend-guard: auto-downgrade sonnet→haiku when ceiling approached
SPEND_GUARD_CEILING_USD=5    # per-session ceiling (default $5)
SPEND_GUARD=on               # set "off" to disable

# Context budget pre-flight (blocks spawns that would overflow)
SPAWN_BUDGET_DEFAULT_CONTEXT=50000  # warning threshold (tokens)
SPAWN_BUDGET_HARD=on                # hard-block spawns exceeding 1.6x

# Query OpenTelemetry trace spans
pnpm run trace:query

# DLQ health (dead-letter queue for failed hook events)
pnpm run dlq:summary
```

--------------------------------

### Copy Agent Studio Configuration (PowerShell)

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Clone or download the Agent Studio repository and copy the configuration files to your project directory using Windows PowerShell commands.

```powershell
Copy-Item -Path ".claude" -Destination "C:\path\to\your\project\.claude" -Recurse
Copy-Item -Path ".claude\CLAUDE.md" -Destination "C:\path\to\your\project\CLAUDE.md"
```

--------------------------------

### Run Agent Integration Tests

Source: https://github.com/oimiragieo/agent-studio/blob/main/GETTING_STARTED.md

Execute integration tests for agent framework and ship readiness. These commands verify the end-to-end functionality of your agents.

```bash
node .claude/tools/verify-agent-integration.mjs --workflow-id agent-integration-v1-<YYYYMMDD-HHMMSS>
```

```bash
pnpm ship-readiness:headless:json
```

```bash
node .claude/tools/verify-ship-readiness.mjs --workflow-id ship-readiness-v1-<YYYYMMDD-HHMMSS> --json
```

--------------------------------

### Implement Async Renderer with Streaming

Source: https://github.com/oimiragieo/agent-studio/blob/main/scripts/channels/PLAN.md

Convert the renderer from synchronous `execSync` to asynchronous `execFile` with stdout piping. This allows for real-time streaming of output chunks.

```javascript
// renderer.cjs
async renderStream(event, onChunk) {
  // Spawn claude -p as child process
  // Pipe stdout line by line
  // Call onChunk(accumulatedText) every ~500ms
  // Return final text on completion
}
```