Agency Commands

Composable workflow commands for automating the complete software development lifecycle.

Command Namespace

All agency commands use the agency: prefix to avoid conflicts:

• /agency:work instead of /work
• /agency:plan instead of /plan

Available Commands

Core Workflow Commands

/agency:work

Work on any issue (auto-detects GitHub/Jira)

Usage:

/agency:work 123              # Work on issue #123
/agency:work next             # Work on next available issue
/agency:work https://...      # Work on issue from URL

Flow: Fetch → Plan → Review Plan → Implement → Test → Review → PR

Agents Used: Orchestrator → Plan → Specialist Reviewer → Specialist Coder → Reality Checker

---

/agency:plan

Create implementation plan only (no execution)

Usage:

/agency:plan 123              # Plan for issue #123
/agency:plan "add dark mode"  # Plan from description

Flow: Analyze → Research → Plan → Review → Present

Agents Used: Orchestrator → Explore → Plan → Specialist Reviewer

---

/agency:implement

Implement from existing plan

Usage:

/agency:implement <plan-file>

Flow: Load Plan → Implement → Test → Review → Commit

Agents Used: Orchestrator → Specialist Coder → Reality Checker

---

GitHub Workflow Commands

/agency:sprint

Implement entire GitHub sprint

Usage:

/agency:sprint                    # Current sprint
/agency:sprint milestone-name     # Specific milestone

Flow: Fetch Issues → Prioritize → Find Parallel → Execute Batches → Monitor

Agents Used: Orchestrator → Sprint Prioritizer → Multiple specialists in parallel

---

/agency:triage

Triage GitHub issues

Usage:

/agency:triage                  # Triage open issues
/agency:triage label:bug        # Triage specific label

Flow: Fetch Issues → Analyze → Categorize → Prioritize → Update Labels

Agents Used: Orchestrator → Senior Project Manager

---

/agency:parallel

Find parallelizable GitHub issues

Usage:

/agency:parallel              # Find all parallelizable work
/agency:parallel sprint       # Only in current sprint

Flow: Fetch Issues → Analyze Dependencies → Group Independent → Report

Agents Used: Orchestrator → Senior Project Manager

---

/agency:worktree

Create git worktree for issue

Usage:

/agency:worktree 123          # Create worktree for issue #123

Flow: Fetch Issue → Create Branch → Create Worktree → Setup Environment

---

Jira Workflow Commands

/agency:sprint

Implement entire Jira sprint

Usage:

/agency:sprint                # Current sprint
/agency:sprint SPRINT-123     # Specific sprint

Flow: Fetch Issues → Prioritize → Find Parallel → Execute Batches → Monitor

---

/agency:triage

Triage Jira issues

Usage:

/agency:triage               # Triage backlog
/agency:triage project-key   # Specific project

Flow: Fetch Issues → Analyze → Categorize → Prioritize → Update Jira

---

Quality & Review Commands

/agency:review

Comprehensive code review

Usage:

/agency:review                    # Review current changes
/agency:review PR-123             # Review specific PR

Flow: Security Review || Quality Review || Performance Review || Test Review → Synthesis

Agents Used: Legal Compliance Checker + Reality Checker + Performance Benchmarker + Test Results Analyzer (parallel)

---

/agency:test

Generate comprehensive tests

Usage:

/agency:test                      # Test current changes
/agency:test <file-path>          # Test specific file

Flow: Analyze Code → Generate Unit Tests → Generate Integration Tests → Generate E2E Tests → Review Coverage

Agents Used: Orchestrator → Testing Specialist

---

/agency:security

Security audit workflow

Usage:

/agency:security                  # Audit current changes
/agency:security --full           # Full codebase audit

Flow: Vulnerability Scan → Code Review → Dependency Check → Report

Agents Used: Legal Compliance Checker

---

Optimization Commands

/agency:optimize

Performance optimization workflow

Usage:

/agency:optimize                  # Optimize current code
/agency:optimize --profile        # Profile first, then optimize

Flow: Profile → Identify Bottlenecks → Optimize → Benchmark → Verify

Agents Used: Performance Benchmarker → Backend/Frontend specialist

---

/agency:refactor

Refactoring workflow

Usage:

/agency:refactor <file-or-pattern>

Flow: Analyze → Plan Refactoring → Execute → Test → Review

Agents Used: Senior Developer → Reality Checker

---

Documentation Commands

/agency:document

Generate documentation

Usage:

/agency:document                  # Document current changes
/agency:document <file-path>      # Document specific file
/agency:document --api            # Generate API docs

Flow: Analyze Code → Generate Docs → Review → Format

---

/agency:adr

Create Architecture Decision Record

Usage:

/agency:adr "use PostgreSQL for persistence"

Flow: Context → Decision → Consequences → Document

---

Command Implementation Status

Phase 1 (Initial Release)

• /agency:work - Core workflow command
• /agency:plan - Planning only
• /agency:review - Code review

Phase 2

• /agency:sprint - Sprint automation
• /agency:triage - Issue triage
• /agency:parallel - Parallel work finder
• /agency:test - Test generation

Phase 3

• /agency:implement - From existing plan
• /agency:optimize - Performance optimization
• /agency:refactor - Refactoring workflow
• /agency:document - Documentation generation

Phase 4

• /agency:worktree - Worktree creation
• /agency:security - Security audit
• /agency:adr - ADR creation

Command Design Principles

1. Composability

Commands should work together:

/agency:plan 123
# Review plan, make changes
/agency:implement plan.md

2. Intelligent Defaults

Commands should work without configuration:

/agency:work next    # Auto-detects next issue

3. Progressive Enhancement

Commands support flags for advanced usage:

/agency:work 123 --auto-approve --model opus --no-tests

4. Consistent Patterns

All commands follow similar structure:

• Fetch/Analyze phase
• Planning/Strategy phase
• Execution phase
• Verification phase
• Delivery phase

5. Checkpoint Strategy

Commands pause for user approval at key points:

• After planning (before implementation)
• Before destructive operations
• When ambiguity detected

Outstanding Design Questions

See enhancements/outstanding-questions/ for detailed design questions about:

• q1-command-priority-selection.md - Which commands to implement first
• q2-interactive-vs-automated-commands.md - Interactive vs automated behavior
• q3-command-argument-patterns.md - Argument format support

Creating New Commands

To create a new command:

1. Create commands/agency-{name}.md
2. Add frontmatter:

---
description: Brief description of what this command does
argument-hint: Hint text shown to user
allowed-tools: [Read, Write, Edit, Bash, Task]
---

3. Write instructions FOR Claude (not TO user)
4. Reference relevant skills and agents
5. Define phases and checkpoints
6. Include error handling

See existing commands for examples.

Related Documentation

• skills/agency-workflow-patterns/ - Orchestration patterns
• skills/github-workflow-best-practices/ - GitHub workflows
• agents/orchestrator.md - Orchestrator agent
• enhancements/outstanding-questions/ - Design decisions