Work Packages: Autonomous Multi-Agent Orchestration Research

Inputs: Design documents from /kitty-specs/019-autonomous-multi-agent-orchestration-research/ Prerequisites: plan.md (research methodology), spec.md (research questions), data-model.md (schema templates), quickstart.md (CLI testing guide)

Mission: Research (no code implementation)

Organization: Fine-grained subtasks (Txxx) roll up into work packages (WPxx). Each work package produces research artifacts that feed into synthesis phases.

Parallelization Note: WP01-WP05 are fully independent and can run simultaneously. WP06-WP08 have dependencies and must run sequentially after their prerequisites complete.

Subtask Format: [Txxx] [P?] Description

  • [P] indicates the subtask can proceed in parallel (different agents/concerns).
  • File paths reference research/XX-agent-name.md for individual findings.

Work Package WP01: Research Known CLI Agents (Priority: P1) [P]

Goal: Investigate the 4 agents with known CLI tools: Claude Code, GitHub Codex, OpenCode, Amazon Q. Independent Test: Each agent has a completed research file with working CLI commands verified locally. Prompt: tasks/WP01-research-known-cli-agents.md

Included Subtasks

  • ✅ T001 [P] Research Claude Code CLI - docs, install, claude --help, test invocation
  • ✅ T002 [P] Research GitHub Codex CLI - docs, install, codex --help, test invocation
  • ✅ T003 [P] Research OpenCode CLI - docs, install, opencode --help, test invocation
  • ✅ T004 [P] Research Amazon Q CLI - docs, install, q --help, test invocation
  • ✅ T005 Write research files: research/01-claude-code.md, research/08-github-codex.md, research/06-opencode.md, research/12-amazon-q.md

Implementation Notes

  • Follow research template from plan.md for each agent
  • All 4 agents can be researched in parallel
  • Must include: CLI availability, task specification method, completion detection, parallel constraints
  • Local testing required: run --help, --version, and basic prompt

Parallel Opportunities

  • All 4 agents are independent - T001-T004 can run simultaneously

Dependencies

  • None (starting package)

Risks & Mitigations

  • CLI may require paid subscription → document free tier limitations
  • Auth tokens needed → document setup without storing secrets

Work Package WP02: Research Cursor CLI (Priority: P1 - User Priority) [P]

Goal: Specifically investigate Cursor's CLI capabilities as requested by user. Independent Test: Cursor CLI documented with working invocation example or confirmed as IDE-only. Prompt: tasks/WP02-research-cursor-cli.md

Included Subtasks

  • ✅ T006 Search for Cursor CLI documentation and installation
  • ✅ T007 Check Cursor.app for embedded CLI tools (macOS: /Applications/Cursor.app/Contents/Resources/)
  • ✅ T008 Test cursor --help or equivalent command
  • ✅ T009 Document headless invocation method if available
  • ✅ T010 Write research file: research/04-cursor.md

Implementation Notes

  • User specifically requested finding Cursor's CLI
  • Check if Cursor has shell command integration like VS Code (cursor .)
  • Investigate agent mode capabilities vs basic editor commands
  • If CLI exists, document full task specification method

Parallel Opportunities

  • T006-T009 are sequential investigation steps

Dependencies

  • None (independent research)

Risks & Mitigations

  • Cursor may be IDE-only → document API alternatives if available
  • CLI may be undocumented → test empirically and note unofficial status

Work Package WP03: Research IDE-Primary Agents (Priority: P2) [P]

Goal: Investigate GitHub Copilot and Windsurf (Codeium) for headless CLI options. Independent Test: Both agents have completed research files documenting CLI availability or IDE-only status. Prompt: tasks/WP03-research-ide-agents.md

Included Subtasks

  • ✅ T011 [P] Research GitHub Copilot - check for gh copilot extension, API access
  • ✅ T012 [P] Research Windsurf/Codeium - check for codeium CLI, language server headless mode
  • ✅ T013 Document any headless workarounds (API calls, extension CLIs)
  • ✅ T014 Write research files: research/02-github-copilot.md, research/07-windsurf.md

Implementation Notes

  • Both are primarily IDE extensions, likely limited headless support
  • Check GitHub CLI (gh) for Copilot extension capabilities
  • Codeium has language server that might run headless

Parallel Opportunities

  • T011 and T012 can run simultaneously

Dependencies

  • None (independent research)

Risks & Mitigations

  • May be IDE-only → document and move on
  • API-only access may have different rate limits

Work Package WP04: Research Cloud/API Agents (Priority: P2) [P]

Goal: Investigate Google Gemini and Qwen Code for CLI/API access. Independent Test: Both agents have completed research files with CLI or API invocation methods. Prompt: tasks/WP04-research-cloud-agents.md

Included Subtasks

  • ✅ T015 [P] Research Google Gemini - check for gemini CLI, gcloud ai commands, API SDK
  • ✅ T016 [P] Research Qwen Code - check Alibaba Cloud CLI, DashScope API, any standalone tools
  • ✅ T017 Document cloud authentication requirements for each
  • ✅ T018 Write research files: research/03-google-gemini.md, research/05-qwen-code.md

Implementation Notes

  • Both likely require cloud SDK or direct API calls
  • Check for official CLI wrappers or third-party tools
  • Document API endpoint structure if no CLI exists

Parallel Opportunities

  • T015 and T016 can run simultaneously

Dependencies

  • None (independent research)

Risks & Mitigations

  • May require cloud account setup → document requirements
  • API-only may need wrapper script for orchestration

Work Package WP05: Research VS Code Extensions (Priority: P2) [P]

Goal: Investigate Kilocode, Augment Code, and Roo Cline for headless capabilities. Independent Test: All 3 agents have completed research files documenting capabilities. Prompt: tasks/WP05-research-vscode-extensions.md

Included Subtasks

  • ✅ T019 [P] Research Kilocode - check for CLI, API, extension command interface
  • ✅ T020 [P] Research Augment Code - check for CLI, API, headless mode
  • ✅ T021 [P] Research Roo Cline - check Cline project for CLI, fork differences
  • ✅ T022 Document VS Code extension command patterns if applicable
  • ✅ T023 Write research files: research/09-kilocode.md, research/10-augment-code.md, research/11-roo-cline.md

Implementation Notes

  • All three are VS Code extensions - may share patterns
  • Roo Cline is a fork of Cline - check original project for CLI
  • Look for @command patterns or task file conventions

Parallel Opportunities

  • T019, T020, T021 can run simultaneously

Dependencies

  • None (independent research)

Risks & Mitigations

  • Extensions typically IDE-only → document and note for orchestration limitations
  • May need VS Code extension host to run → document as limitation

Work Package WP06: Synthesize CLI Capability Matrix (Priority: P1)

Goal: Consolidate all research findings into comprehensive capability matrix. Independent Test: research.md updated with complete matrix and orchestration assessment. Prompt: tasks/WP06-synthesize-capability-matrix.md

Included Subtasks

  • ✅ T024 Review all 12 agent research files for completeness
  • ✅ T025 Build CLI capability matrix (agent × capability columns)
  • ✅ T026 Identify orchestration-ready agents (can participate fully)
  • ✅ T027 Identify partially-capable agents (need workarounds)
  • ✅ T028 Identify non-capable agents (cannot participate)
  • ✅ T029 Update research.md with consolidated findings
  • ✅ T030 Verify quality gates: QG-001 (≥6 CLI agents), QG-002 (Cursor), QG-003 (sources), QG-004 (parallel constraints)

Implementation Notes

  • Cross-reference all research files
  • Ensure every cell in matrix has data or explicit "N/A"
  • Include source links for all findings

Parallel Opportunities

  • None - sequential synthesis

Dependencies

  • Depends on: WP01, WP02, WP03, WP04, WP05 (all research complete)

Risks & Mitigations

  • Missing data → flag incomplete research for follow-up
  • Conflicting findings → verify with local testing

Work Package WP07: Design Agent Orchestration Config (Priority: P2)

Goal: Propose concrete configuration schema for agent preferences based on findings. Independent Test: data-model.md updated with realistic field values and working example config. Prompt: tasks/WP07-design-orchestration-config.md

Included Subtasks

  • ✅ T031 Review capability matrix for config requirements
  • ✅ T032 Refine AgentProfile schema with real data from research
  • ✅ T033 Design OrchestratorConfig with practical defaults
  • ✅ T034 Document fallback strategies based on agent availability
  • ✅ T035 Handle single-agent edge case (same agent for impl & review)
  • ✅ T036 Update data-model.md with concrete examples
  • ✅ T037 Create sample .kittify/agents.yaml config file

Implementation Notes

  • Schema must accommodate all CLI-capable agents discovered
  • Include realistic rate limits and constraints from research
  • Provide sensible defaults based on findings

Parallel Opportunities

  • T032-T35 can be developed incrementally

Dependencies

  • Depends on: WP06 (need capability matrix to design realistic config)

Risks & Mitigations

  • Few CLI-capable agents → simplify config, emphasize single-agent mode

Work Package WP08: Final Report & Recommendations (Priority: P1)

Goal: Write executive summary, feasibility assessment, and architecture recommendations. Independent Test: Complete research report ready for stakeholder review. Prompt: tasks/WP08-final-report.md

Included Subtasks

  • ✅ T038 Write executive summary in research.md
  • ✅ T039 Document feasibility assessment: can autonomous orchestration work?
  • ✅ T040 Identify minimum viable agent set for orchestration
  • ✅ T041 Propose architecture approach for orchestrator implementation
  • ✅ T042 Document gaps and future research needs
  • ✅ T043 Final quality gate verification
  • ✅ T044 Update all documentation with cross-references

Implementation Notes

  • Focus on actionable recommendations
  • Be honest about limitations discovered
  • Provide clear next steps for implementation phase

Parallel Opportunities

  • None - final synthesis

Dependencies

  • Depends on: WP06 (capability matrix), WP07 (config schema)

Risks & Mitigations

  • Findings may be disappointing → document honestly, suggest alternatives

Dependency & Execution Summary

WP01 (Known CLI)     ─┐
WP02 (Cursor)        ─┼─→ WP06 (Synthesis) ─→ WP07 (Config) ─→ WP08 (Report)
WP03 (IDE Agents)    ─┤
WP04 (Cloud Agents)  ─┤
WP05 (VS Code Ext)   ─┘
  • Phase 1 (Parallel): WP01, WP02, WP03, WP04, WP05 - all independent
  • Phase 2: WP06 - depends on Phase 1
  • Phase 3: WP07 - depends on WP06
  • Phase 4: WP08 - depends on WP06, WP07

Parallelization: 5 research WPs can run simultaneously, dramatically reducing research time.

MVP Scope: WP01 + WP02 + WP06 (Known CLI agents + Cursor + basic synthesis)

Subtask Index (Reference)

Subtask IDSummaryWork PackagePriorityParallel?
T001Research Claude Code CLIWP01P1Yes
T002Research GitHub Codex CLIWP01P1Yes
T003Research OpenCode CLIWP01P1Yes
T004Research Amazon Q CLIWP01P1Yes
T005Write Known CLI research filesWP01P1No
T006Search Cursor CLI docsWP02P1No
T007Check Cursor.app CLI toolsWP02P1No
T008Test cursor --helpWP02P1No
T009Document Cursor headless methodWP02P1No
T010Write Cursor research fileWP02P1No
T011Research GitHub CopilotWP03P2Yes
T012Research Windsurf/CodeiumWP03P2Yes
T013Document headless workaroundsWP03P2No
T014Write IDE agents research filesWP03P2No
T015Research Google GeminiWP04P2Yes
T016Research Qwen CodeWP04P2Yes
T017Document cloud auth requirementsWP04P2No
T018Write cloud agents research filesWP04P2No
T019Research KilocodeWP05P2Yes
T020Research Augment CodeWP05P2Yes
T021Research Roo ClineWP05P2Yes
T022Document VS Code patternsWP05P2No
T023Write VS Code ext research filesWP05P2No
T024Review all research filesWP06P1No
T025Build capability matrixWP06P1No
T026Identify orchestration-ready agentsWP06P1No
T027Identify partially-capable agentsWP06P1No
T028Identify non-capable agentsWP06P1No
T029Update research.mdWP06P1No
T030Verify quality gatesWP06P1No
T031Review matrix for config needsWP07P2No
T032Refine AgentProfile schemaWP07P2No
T033Design OrchestratorConfigWP07P2No
T034Document fallback strategiesWP07P2No
T035Handle single-agent edge caseWP07P2No
T036Update data-model.mdWP07P2No
T037Create sample agents.yamlWP07P2No
T038Write executive summaryWP08P1No
T039Feasibility assessmentWP08P1No
T040Minimum viable agent setWP08P1No
T041Orchestrator architecture proposalWP08P1No
T042Document gaps & future researchWP08P1No
T043Final quality gate verificationWP08P1No
T044Update cross-referencesWP08P1No

<!-- status-model:start -->

Canonical Status (Generated)

<!-- status-model:end -->

  • WP01: done
  • WP02: done
  • WP03: done
  • WP04: done
  • WP05: done
  • WP06: done
  • WP07: done
  • WP08: done