Specialized Agents

Specialized agents

🌍 English narrative is auto-generated on release. See the French version for the current full descriptions.

Beyond the main pipeline (Planner / Coder / QA), WorkPilot AI ships a library of specialized agents invokable from the Kanban, command palette, or CLI. The list below is regenerated automatically from apps/backend/prompts/.

Agent	Prompt file	Category
architecture_reviewer	apps/backend/prompts/architecture_reviewer.md	Analysis
architecture_visualizer	apps/backend/prompts/architecture_visualizer.md	Analysis
breaking_change_detector	apps/backend/prompts/breaking_change_detector.md	Migration
browser_agent	apps/backend/prompts/browser_agent.md	Browser
code_migration	apps/backend/prompts/code_migration.md	Migration
coder	apps/backend/prompts/coder.md	Pipeline
coder_recovery	apps/backend/prompts/coder_recovery.md	Self-Healing
competitor_analysis	apps/backend/prompts/competitor_analysis.md	Roadmap
complexity_assessor	apps/backend/prompts/complexity_assessor.md	Pipeline
documentation_agent	apps/backend/prompts/documentation_agent.md	Documentation
environment_cloner	apps/backend/prompts/environment_cloner.md	Migration
followup_planner	apps/backend/prompts/followup_planner.md	Followup
ideation_code_improvements	apps/backend/prompts/ideation_code_improvements.md	Ideation
ideation_code_quality	apps/backend/prompts/ideation_code_quality.md	Ideation
ideation_documentation	apps/backend/prompts/ideation_documentation.md	Ideation
ideation_performance	apps/backend/prompts/ideation_performance.md	Ideation
ideation_security	apps/backend/prompts/ideation_security.md	Ideation
ideation_ui_ux	apps/backend/prompts/ideation_ui_ux.md	Ideation
incident_cicd_analyzer	apps/backend/prompts/incident_cicd_analyzer.md	Self-Healing
incident_proactive_analyzer	apps/backend/prompts/incident_proactive_analyzer.md	Self-Healing
incident_production_responder	apps/backend/prompts/incident_production_responder.md	Self-Healing
insight_extractor	apps/backend/prompts/insight_extractor.md	Analysis
intent_templates	apps/backend/prompts/intent_templates.md	Followup
learning_analyzer	apps/backend/prompts/learning_analyzer.md	Analysis
multi_repo_planner	apps/backend/prompts/multi_repo_planner.md	Multi-repo
performance_profiler	apps/backend/prompts/performance_profiler.md	Analysis
planner	apps/backend/prompts/planner.md	Pipeline
qa_fixer	apps/backend/prompts/qa_fixer.md	Pipeline
qa_reviewer	apps/backend/prompts/qa_reviewer.md	Pipeline
roadmap_discovery	apps/backend/prompts/roadmap_discovery.md	Roadmap
roadmap_features	apps/backend/prompts/roadmap_features.md	Roadmap
spec_critic	apps/backend/prompts/spec_critic.md	Other
spec_gatherer	apps/backend/prompts/spec_gatherer.md	Other
spec_quick	apps/backend/prompts/spec_quick.md	Other
spec_researcher	apps/backend/prompts/spec_researcher.md	Other
spec_writer	apps/backend/prompts/spec_writer.md	Other
validation_fixer	apps/backend/prompts/validation_fixer.md	Other
issue_analyzer	apps/backend/prompts/github/issue_analyzer.md	GitHub
issue_triager	apps/backend/prompts/github/issue_triager.md	GitHub
pr_ai_triage	apps/backend/prompts/github/pr_ai_triage.md	GitHub
pr_codebase_fit_agent	apps/backend/prompts/github/pr_codebase_fit_agent.md	GitHub
pr_finding_validator	apps/backend/prompts/github/pr_finding_validator.md	GitHub
pr_followup	apps/backend/prompts/github/pr_followup.md	GitHub
pr_followup_comment_agent	apps/backend/prompts/github/pr_followup_comment_agent.md	GitHub
pr_followup_newcode_agent	apps/backend/prompts/github/pr_followup_newcode_agent.md	GitHub
pr_followup_orchestrator	apps/backend/prompts/github/pr_followup_orchestrator.md	GitHub
pr_followup_resolution_agent	apps/backend/prompts/github/pr_followup_resolution_agent.md	GitHub
pr_logic_agent	apps/backend/prompts/github/pr_logic_agent.md	GitHub
pr_orchestrator	apps/backend/prompts/github/pr_orchestrator.md	GitHub
pr_parallel_orchestrator	apps/backend/prompts/github/pr_parallel_orchestrator.md	GitHub
pr_quality_agent	apps/backend/prompts/github/pr_quality_agent.md	GitHub
pr_reviewer	apps/backend/prompts/github/pr_reviewer.md	GitHub
pr_security_agent	apps/backend/prompts/github/pr_security_agent.md	GitHub
pr_structural	apps/backend/prompts/github/pr_structural.md	GitHub
pr_template_filler	apps/backend/prompts/github/pr_template_filler.md	GitHub

Auto-generated from apps/backend/prompts/ · 55 agents.

Specialized Agents

Beyond the core Planner → Coder → QA Reviewer/Fixer pipeline that drives every build, WorkPilot AI ships a set of specialized agents that handle narrower jobs — either standalone or as subagents spawned by the Coder during implementation.

Agent	Role
Planner	Complexity assessment, phased subtask breakdown, dependency analysis
Coder	Context-aware implementation with parallel subagent spawning
QA Reviewer / Fixer	Acceptance criteria validation and automated issue resolution
Test Generator	Unit and integration test generation, streaming code live as it's written
Flaky Test Detective	Parses test reports (JUnit XML, .NET TRX, …) to surface intermittently failing tests
Refactorer	Safe code refactoring with pattern detection and API migration
Documenter	README, API docs, and architecture documentation generation
Migration Agent	Framework and library migration with breaking change detection
Release Coordinator	Orchestrates release readiness, changelog, and version bumps
Memory Manager	Graphiti-based knowledge graph management across sessions

These agents are backed by a growing library of task-specific prompts under apps/backend/prompts/ — covering ideation (performance, security, code quality, UI/UX), incident response, architecture review, and more — so each specialized task gets a purpose-built prompt rather than a generic one.

All agent work happens inside isolated git worktrees, keeping the main branch untouched until a human reviews and merges the result. For workloads that benefit from running several agents at once, WorkPilot AI supports up to 12 parallel agent terminals, along with a Mission Control dashboard for monitoring live status, token consumption, and per-agent reasoning across the fleet. Higher-level orchestration features — like the Consensus Arbiter, which reconciles conflicting findings from security scans and QA reviewers, and Agent Coach, which suggests cheaper models or better effort levels based on real usage — sit on top of this agent layer to keep multi-agent runs coherent and cost-aware.

---

➡️ Continue: Integrations