| name | orchestrate |
| description | God-mode orchestrator. Takes a GitHub issue or PR and handles the ENTIRE workflow end-to-end — branch creation, implementation, testing, commits, PR creation, OR exhaustive PR review with parallel agents. Use for complete feature delivery or deep PR audits. |
| argument-hint | [issue-number or PR branch/URL] [--review] [--auto] |
| disable-model-invocation | true |
Orchestrator: End-to-End Workflow
Two operating modes:
Mode A — Feature Delivery (default)
Takes a GitHub issue and handles the entire pipeline:
- Fetch and understand the issue
- Create a git branch
- Plan implementation
- Implement across all files with maximum parallelism
- Build-review-fix loop until clean
- Commit, push, create PR
Mode B — PR Review (--review)
Takes a PR branch name or URL and performs an exhaustive audit:
- Fetch all commits and changed files
- Spawn parallel review agents across every domain
- Report findings with file:line references and fix suggestions
Parallelization Philosophy
Use as many parallel agents as needed to complete tasks in the fastest, most efficient, and highest-quality way possible. Spawn agents aggressively for independent work streams.
Worker Coordination Protocol
- File-based coordination — Each agent works on independent files. Never assign two agents to the same file.
- Cross-agent synthesis — The orchestrator reads deliverables and feeds relevant context between agents. Workers never talk directly.
- Quality gates with push-back — No agent's work is marked done until deliverables are reviewed.
- Stall detection — If an agent is stuck, reassess and push with specific unblocking instructions.
- Convergence criteria — When parallel agents start producing overlapping concerns, collapse to sequential.
Step-by-Step: Feature Delivery (Mode A)
Step 1: Fetch Issue
gh issue view NUMBER --repo vultisig/vultisig-ios
Extract title, description, labels, acceptance criteria.
Step 2: Git Branch Setup
git checkout main && git pull
git checkout -b feature/ISSUE-description
Step 3: Plan Implementation
- Understand requirements and acceptance criteria
- Map architecture: which files/modules are affected
- Identify dependencies between files
- Create parallelization strategy
- Present to user (unless
--auto)
Step 4: Implement (PARALLEL where possible)
Split implementation across independent modules/files:
- Agent A: Core logic (ViewModels, Services)
- Agent B: UI components (Views)
- Agent C: Localization (all 7 Localizable.strings files)
- Agent D: Tests
Follow ALL .claude/rules/ strictly.
Step 5: Build-Review-Fix Loop
Persistent loop — do NOT stop until ALL gates pass:
while true:
1. Run swiftlint lint --config VultisigApp/.swiftlint.yml VultisigApp/
2. If fails → fix issues, go to 1
3. Run xcodebuild (if available)
4. If fails → fix issues, go to 1
5. All gates pass → break
Step 6: Commit
git add <specific-files>
git commit -m "$(cat <<'EOF'
feat(scope): imperative description
- Detail 1
- Detail 2
Co-Authored-By: Claude <noreply@anthropic.com>
EOF
)"
Step 7: Push + Create PR
git push -u origin $(git branch --show-current)
gh pr create --title "feat(scope): description" --body "$(cat <<'EOF'
## Summary
- What and why
## Test Plan
- [ ] SwiftLint passes
- [ ] Build succeeds
- [ ] Manual verification
## Related
- Closes #ISSUE_NUMBER
Co-Authored-By: Claude <noreply@anthropic.com>
EOF
)" --draft
PR Review (Mode B)
Stage 1 — Gather PR Context (PARALLEL)
- Agent A: Fetch all commits, diff, and changed file list
- Agent B: Read project rules and CLAUDE.md for standards
- Agent C: Identify all domains touched (chain logic, UI, state, etc.)
Stage 2 — Deep Review (PARALLEL, one agent per domain)
- Architecture Agent: MVVM compliance, business logic placement, service injection
- Code Quality Agent: DRY violations, dead code, force unwraps, SwiftLint compliance
- Concurrency Agent: async/await correctness, MainActor usage, data races
- UI Agent: Theme token usage, PrimaryButton, Screen component, .foregroundStyle()
- Localization Agent: All 7 files updated, camelCase keys, alphabetical order
- Security Agent: Key material exposure, credential leaks, TSS boundary violations
Stage 3 — Synthesize & Report (SEQUENTIAL)
- Merge findings, deduplicate, prioritize: critical > important > nitpick
- Present report with file:line references
Pre-Submit Gate