AI Agent Integration

Conduct seamlessly integrates with popular AI coding tools, providing them with custom commands and instructions for managing your project specifications.

Supported AI Agents

Cursor

Command Location: .cursor/commands/

Cursor commands are available via the command palette and slash commands:

• /feature - Create a new feature specification
• /change - Create a change specification

Files Created:

• .cursor/commands/feature.md
• .cursor/commands/change.md

Claude Code

Command Location: .claude/commands/conduct/

Claude commands are available through custom slash commands:

• /feature - Create a new feature specification
• /change - Create a change specification

Files Created:

• .claude/commands/conduct/feature.md
• .claude/commands/conduct/change.md
• CLAUDE.md - Root-level instructions automatically loaded

Special Features:

• Claude reads CLAUDE.md automatically on project load
• Instructions are injected into Claude’s context without manual intervention

Warp

Command Location: .warp/commands/conduct/

Warp commands are available in Agent Mode:

• /feature - Create a new feature specification
• /change - Create a change specification

Files Created:

• .warp/commands/conduct/feature.md
• .warp/commands/conduct/change.md
• WARP.md - Root-level instructions for Warp Agent Mode

Special Features:

• Warp reads WARP.md automatically in Agent Mode
• Commands are optimized for terminal-based workflow

Factory

Command Location: .factory/commands/

Factory commands integrate directly with Factory’s command system:

• /feature - Create a new feature specification
• /change - Create a change specification

Files Created:

• .factory/commands/feature.md
• .factory/commands/change.md

OpenCode

Command Location: .opencode/command/

OpenCode commands are available through custom slash commands:

• /feature - Create a new feature specification
• /change - Create a change specification

Files Created:

• .opencode/command/feature.md
• .opencode/command/change.md

GitHub Copilot

Command Location: .github/prompts/

GitHub Copilot uses prompt files for custom instructions:

Files Created:

• .github/prompts/conduct-feature.prompt.md
• .github/prompts/conduct-change.prompt.md

Usage: Reference these prompts in GitHub Copilot Chat or use them as context for code generation.

How Agent Integration Works

1. Command Files

Each agent receives markdown files containing:

• Command syntax and usage
• Step-by-step instructions for creating specs
• Templates for feature and change specifications
• Naming conventions and best practices
• Issue tracker integration guidelines

2. Automatic Context Loading

Some agents (Claude, Warp) automatically load root-level instruction files:

• CLAUDE.md - Loaded by Claude on project open
• WARP.md - Loaded by Warp in Agent Mode

These files contain:

• Overview of Conduct structure
• Links to agent commands
• Project-specific conventions

3. Dynamic Content

Agent commands include dynamic placeholders that get replaced at runtime:

• {{ISSUE_TRACKER_LINKS}} - Links to configured issue trackers
• {{FEATURE_PATH}} - Current feature context
• {{VERSION}} - Current spec version

Using Agent Commands

Creating a Feature

1. Open your AI agent (Cursor, Claude, Warp, etc.)
2. Use the /feature command:

   /feature User authentication system
3. The agent will:
   • Create conduct/features/user-authentication-system/
   • Generate spec.v0.md with proper structure
   • Create nested features/ and changes/ directories
   • Link to issue tracker if configured

Creating a Change

1. Open your AI agent
2. Use the /change command:

   /change Add OAuth2 support
3. The agent will:
   • Prompt for parent feature (if multiple exist)
   • Create change spec in feature’s changes/ directory
   • Increment parent feature version (v0 → v1)
   • Update track.json with new version

Nested Features

To create a nested feature:

1. Navigate to a feature directory in your codebase
2. Use /feature command within that context:

   /feature OAuth integration
3. The agent creates a nested feature:

   conduct/features/authentication/features/oauth/spec.v0.md

Agent-Specific Tips

Cursor

• Commands appear in the command palette (Cmd+K / Ctrl+K)
• Use in chat mode for interactive feature creation
• Commands respect current file/directory context

Claude Code

• CLAUDE.md is automatically loaded on project open
• No manual command registration needed
• Commands work in both chat and inline modes

Warp

• Commands are optimized for terminal workflows
• WARP.md provides context in Agent Mode
• Use commands directly in chat interface

Factory

• Commands integrate with Factory’s workflow system
• Supports Factory’s multi-step command execution
• Respects Factory’s project structure conventions

OpenCode

• Commands follow OpenCode’s slash command conventions
• Integrates with OpenCode’s project awareness
• Supports OpenCode’s multi-file editing

GitHub Copilot

• Use prompt files as context in Copilot Chat
• Reference prompts: @conduct-feature, @conduct-change
• Works with Copilot’s code completion

Customizing Agent Instructions

Upgrading Instructions

Keep agent instructions up to date:

conduct upgrade

This updates all agent command files to the latest version.

Manual Customization

You can manually edit agent command files to:

• Add project-specific conventions
• Include custom templates
• Add additional validation steps
• Modify naming patterns

Note: Running conduct upgrade will overwrite manual changes. Back up customizations before upgrading.

Per-Agent Configuration

Each agent can have different instructions if needed:

1. Edit agent-specific command files
2. Add custom steps or requirements
3. Maintain consistency across agents for team collaboration

Issue Tracker Integration

Conduct integrates with issue trackers to link specs to tickets:

Supported Trackers

• GitHub Issues: https://github.com/org/repo/issues/123
• Jira: https://company.atlassian.net/browse/PROJ-123
• Linear: https://linear.app/company/issue/PROJ-123
• GitLab: https://gitlab.com/org/repo/-/issues/123
• Azure DevOps: https://dev.azure.com/org/project/_workitems/edit/123

How It Works

1. Configure trackers during conduct init
2. Agent commands include tracker links in templates
3. AI agents automatically generate proper links in specs

Example in Spec

## Links

- [GitHub Issue #123](https://github.com/myorg/myrepo/issues/123)
- [Jira Ticket PROJ-456](https://mycompany.atlassian.net/browse/PROJ-456)

Multi-Agent Workflows

Same Project, Multiple Agents

You can use multiple AI agents in the same project:

1. Configure all agents during conduct init
2. Each agent gets its own command files
3. All agents follow the same specification structure
4. Changes made by any agent are compatible with others

Team Collaboration

When working in teams:

1. Commit .cursor/, .claude/, .warp/, etc. to version control
2. Team members use their preferred AI agents
3. Everyone follows the same Conduct conventions
4. Specifications remain consistent regardless of agent used

Switching Between Agents

You can switch between agents seamlessly:

1. All agents read the same conduct/ specifications
2. Commands are equivalent across agents
3. No migration needed when switching

Troubleshooting

Commands Not Appearing

Cursor: Reload window (Cmd+R / Ctrl+R)

Claude: Restart Claude Code editor

Warp: Restart Warp or reload Agent Mode

Other agents: Refer to agent-specific documentation

Commands Not Working

1. Verify agent command files exist
2. Check file permissions
3. Ensure conduct/ directory exists
4. Run conduct upgrade to refresh instructions

Outdated Instructions

Run conduct upgrade to update all agent instructions to the latest version.

Custom Modifications Lost

• Back up custom changes before running conduct upgrade
• Maintain custom instructions in separate files
• Use version control to track changes

Next Steps

• Learn about Architecture
• Read the CLI Reference
• Explore specification structure