Taskforces are pre-configured multi-agent orchestrations that coordinate specialized agents to accomplish specific types of missions. Each taskforce combines a curated set of agents with a custom orchestrator that manages the workflow from start to finish.

Placeholder: Taskforce selection dropdown showing taskforce names, descriptions, and success criteria - Bug Fix, Feature Implementation, Refactoring, etc.
Select a taskforce when creating a new chat to use coordinated multi-agent workflows

What are Taskforces?

A taskforce bundles together:

Specialized Orchestrator: Domain-specific coordination logic that knows how to sequence and parallelize agent work for maximum effectiveness.

Curated Agent Team: A hand-picked set of agents with complementary skills selected specifically for the workflow.

Success Criteria: Clear, measurable conditions that define when the mission is complete, ensuring quality and completeness.

YAML Configuration: Declarative configuration makes taskforces easy to customize, share, and version control.

The orchestrator handles all the complexity of multi-agent coordination - deciding which agents to use, when to run them in parallel versus sequentially, and ensuring success criteria are met before considering the work complete.

When to Use Taskforces

Taskforces excel at well-defined workflows requiring multiple specialized agents:

Use taskforces when:

  • You have a systematic process that benefits from multiple perspectives (research, implementation, review)
  • Quality gates and verification are critical to success
  • Work naturally phases through stages (investigation → planning → implementation → validation)
  • You want proven workflows for common development patterns

Use individual agents when:

  • Working on exploratory or open-ended tasks
  • The work is specialized to a single domain
  • You prefer direct control over the approach
  • The task doesn’t benefit from multiple agent perspectives

Built-in Taskforces

Reliant includes three production-ready taskforces for common development workflows:

Bug Fix Taskforce

Purpose: Systematic bug investigation, resolution, and verification with zero tolerance for unverified fixes.

Agents: research, repro, general, debug_with_me, checklist

Workflow:

  1. Investigation: research analyzes codebase to understand architecture and locate relevant code
  2. Reproduction: repro creates reliable reproduction steps and test cases
  3. Root Cause Analysis: debug_with_me uses debugging tools to identify the exact cause
  4. Fix Implementation: general implements the fix with regression tests
  5. Verification: checklist ensures the fix works and nothing broke

Success Criteria:

  • Bug reliably reproduced with documented steps
  • Root cause identified and clearly explained
  • Fix addresses root cause, not just symptoms
  • Tests added to prevent regression
  • All existing tests still pass
  • No new bugs introduced

Best For: Fixing reported bugs, investigating production issues, ensuring fixes are properly verified before deployment.

Feature Implementation Taskforce

Purpose: End-to-end feature development from requirements analysis through delivery.

Agents: research, planner, general, code_reviewer, checklist, finalizer

Workflow:

  1. Research: research investigates existing patterns and architecture
  2. Planning: planner creates detailed implementation plan with task breakdown
  3. Implementation: general builds the feature incrementally with tests
  4. Quality Review: code_reviewer ensures code quality and best practices
  5. Verification: checklist validates all requirements are met
  6. Finalization: finalizer polishes documentation and prepares for deployment

Success Criteria:

  • All acceptance criteria met and verified
  • Feature works end-to-end with comprehensive testing
  • Code review completed with feedback addressed
  • All tests pass with adequate coverage
  • Documentation complete (code, API, user docs)
  • No regressions in existing functionality
  • Ready for production deployment

Best For: Implementing new features, building complete functionality, projects requiring comprehensive quality assurance.

Refactor Taskforce

Purpose: Safe, systematic code refactoring while maintaining behavior and test coverage.

Agents: research, planner, general, code_reviewer, checklist

Workflow:

  1. Analysis: research identifies code smells and refactoring opportunities
  2. Planning: planner creates incremental refactoring plan with safety checkpoints
  3. Test Coverage: general adds tests to establish behavioral baseline
  4. Refactoring: general executes refactoring in small, safe steps
  5. Review: code_reviewer validates quality improvements
  6. Verification: checklist ensures behavior unchanged and all tests pass

Success Criteria:

  • All tests pass (no behavior changes)
  • Code quality metrics improved
  • Behavior unchanged (verified by tests)
  • No performance degradation
  • Code more maintainable and understandable

Best For: Improving code quality, restructuring code, reducing technical debt while maintaining confidence that nothing breaks.

Using Taskforces

Starting a Chat with a Taskforce

When creating a new chat, select a taskforce from the taskforce dropdown:

  1. Click the New Chat button
  2. Select a taskforce from the Taskforce dropdown
  3. The chat will use that taskforce’s multi-agent coordination
  4. Provide your initial message describing the work

The orchestrator will analyze your request and coordinate the appropriate agents automatically.

Taskforce Workflow Visibility

While a taskforce works:

  • You’ll see which agent is currently active
  • Agent transitions are logged for transparency
  • You can see progress toward success criteria
  • Each agent’s work is clearly separated in the conversation

Switching Between Taskforces

Each chat is bound to a single taskforce (or individual agent) for its lifetime. To use a different taskforce, create a new chat.

This design ensures:

  • Consistent coordination throughout the workflow
  • Clear expectations about how work will be handled
  • Orchestrators can maintain state across the entire mission