Aiwg team

Orchestrate multi-agent teams across AIWG providers natively on Claude Code or emulated via Mission Control elsewhere

install
source · Clone the upstream repo
git clone https://github.com/jmagly/aiwg
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/jmagly/aiwg "$T" && mkdir -p ~/.claude/skills && cp -r "$T/agentic/code/addons/aiwg-utils/skills/team" ~/.claude/skills/jmagly-aiwg-team-ba2640 && rm -rf "$T"
manifest: agentic/code/addons/aiwg-utils/skills/team/SKILL.md
source content

team

You orchestrate multi-agent teams across all AIWG providers. On Claude Code, teams run natively via the Task tool. On all other providers, you emulate team execution through

aiwg mc
(Mission Control). Team definitions live in
agentic/code/frameworks/sdlc-complete/teams/
.

Triggers

Alternate expressions and non-obvious activations (primary phrases are matched automatically from the skill description):

  • "spin up a team" →
    team run
    with named team
  • "who's on the security review team" →
    team info
  • "what teams do I have" →
    team list
  • "run the greenfield team on this task" →
    team run greenfield "<task>"

Trigger Patterns Reference

PatternExampleAction
Run a named team"run the sdlc-review team on the SAD"
aiwg team run sdlc-review "<task>"
List available teams"what teams are available"
aiwg team list
Inspect team composition"who is on the security review team"
aiwg team info security-review
Task-inferred team"review this architecture for security"
aiwg team run security-review "<task>"

Behavior

When triggered:

  1. Parse the subcommand:

    • run <team-name> "<task>"
      — execute team against a task
    • list
      — list all available teams from the manifest
    • info <team-name>
      — show team composition and use cases
  2. Detect provider:

    • Claude Code (
      CLAUDE_CODE_VERSION
      set) → native Task-tool dispatch
    • All other providers → emulate via
      aiwg mc dispatch
  3. Run the appropriate command:

    # Run a team on a task
    aiwg team run sdlc-review "Review the SAD at .aiwg/architecture/sad.md"
    
    # Run with provider override
    aiwg team run security-review "Audit auth module" --provider cursor
    
    # List all teams
    aiwg team list
    
    # Inspect a team
    aiwg team info greenfield
    
  4. Report execution progress:

    • Native (Claude Code): reports inline as agents complete
    • Emulated (mc): reports mission IDs and tracks via
      aiwg mc status

Team Definitions

Teams are defined as JSON files in

agentic/code/frameworks/sdlc-complete/teams/
. Built-in teams:

TeamUse Cases
sdlc-review
Architecture, requirements, and test review
security-review
Threat modeling, security audit, privacy review
api-development
API design, documentation, test generation
full-stack
Feature implementation across layers
greenfield
New project inception and architecture
maintenance
Refactoring, debt reduction, upgrades
migration
Database or platform migration planning

Team Definition Format

# .aiwg/teams/my-team.yaml  (project-local override)
name: Custom Review Team
slug: custom-review
description: Tailored review for domain-specific concerns
agents:
  - agent: security-architect
    role: lead
    responsibilities:
      - Security threat model
      - Auth design review
  - agent: test-architect
    role: contributor
    responsibilities:
      - Test coverage gaps
      - Integration risks
  - agent: technical-writer
    role: reviewer
    responsibilities:
      - Clarity and consistency
dispatch: parallel

Execution Modes

Native (Claude Code)

Agents launch as parallel Task tool calls in a single message. Each agent runs independently with its assigned prompt, and results are collected and synthesized.

Emulated (All Other Providers)

# Steward emulates by dispatching via Mission Control
SESSION=$(aiwg mc start --name "team-run-$(date +%s)")
aiwg mc dispatch $SESSION "security-architect: <task>" --completion "Review complete"
aiwg mc dispatch $SESSION "test-architect: <task>" --completion "Review complete"
aiwg mc watch $SESSION

Examples

Example 1: Run a review team

User: "Run the security review team on the auth module"

Extraction:

team run security-review
on the auth module

Action:

aiwg team run security-review "Review src/auth/ for security vulnerabilities, focusing on token handling and session management"

Response (Claude Code, native):

Launching security-review team (3 agents, parallel)...
  security-architect: reviewing...
  security-auditor:   reviewing...
  privacy-officer:    reviewing...

Results collected. 2 critical findings, 4 recommendations.
Full report: .aiwg/working/team-runs/security-review-20260401-143022.md

Example 2: List teams

User: "What teams do I have available?"

Action:

aiwg team list

Response:

Available teams (7):
  api-development   API design, documentation, test generation
  full-stack        Feature implementation across layers
  greenfield        New project inception and architecture
  maintenance       Refactoring, debt reduction, upgrades
  migration         Database or platform migration planning
  sdlc-review       Architecture, requirements, and test review
  security-review   Threat modeling, security audit, privacy review

Example 3: Inspect team composition

User: "Who is on the sdlc-review team?"

Action:

aiwg team info sdlc-review

Response:

Team: sdlc-review
Description: Architecture, requirements, and test review
Dispatch: parallel

Agents:
  architecture-designer  [lead]        Architecture quality and pattern review
  requirements-analyst   [contributor] Requirements traceability and completeness
  test-architect         [contributor] Testability and coverage gaps
  technical-writer       [reviewer]    Clarity and documentation quality

Use cases:
  - SAD review before Architecture Baseline Milestone
  - Elaboration phase gate check
  - ADR peer review

Clarification Prompts

If the user names a task without a team:

  • "Which team should I run? Available: sdlc-review, security-review, api-development, full-stack, greenfield, maintenance, migration"

If the team name is ambiguous:

  • "Did you mean
    security-review
    or
    sdlc-review
    ?"

References

  • @$AIWG_ROOT/src/cli/handlers/team.ts — Team command handler
  • @$AIWG_ROOT/agentic/code/frameworks/sdlc-complete/teams/ — Built-in team definitions
  • @$AIWG_ROOT/agentic/code/frameworks/sdlc-complete/teams/manifest.json — Team manifest
  • @$AIWG_ROOT/agentic/code/frameworks/sdlc-complete/teams/schema.json — Team definition schema
  • @$AIWG_ROOT/docs/cli-reference.md — CLI reference