AI-powered implementation planning with agent assignments and parallel execution
Invoke this skill in Claude Code:
Takes a solution from /createsolution and produces a phased implementation plan. Three AI models create independent plans, which are synthesized and peer-reviewed to produce a final plan with agent assignments and parallel execution guidance.
Task sequencing, milestone tracking, cross-agent coordination
Infrastructure design, tech stack decisions, integration patterns
OAuth2, firewall rules, SSL/TLS, security hardening
Coding, Docker deployment, database ops, bug fixes
Test specs, CI pipeline, validation, acceptance criteria
The plan identifies which tasks can run in parallel (same group letter) vs. sequential (dependencies). Multiple developers can work on different services simultaneously, while architects plan the next phase.
2-3 paragraph overview of the project scope and approach.
Organized phases with tasks, agents, parallel groups, and dependencies.
Mermaid diagram showing task dependencies and parallel execution paths.
Decisions requiring human input before implementation. Become DECISION cards in /executeplan.
Potential blockers and mitigations identified during planning.
Tasks per agent with parallel execution opportunities for balanced distribution.
{descriptor}.plan.ask.md - Clarified plan ask{descriptor}.plan.claude.md - Claude's plan{descriptor}.plan.gemini.md - Gemini's plan{descriptor}.plan.codex.md - Codex's plan{descriptor}.plan.draft.md - Synthesized draft{descriptor}.plan.gemini-review.md - Gemini peer review{descriptor}.plan.codex-review.md - Codex peer review{descriptor}.plan.final.md - Final plan
# Create implementation plan (reads solution from createsolution)
/createplan Build the stock portfolio tracker
# Output locations:
# Markdown: ~/projects/ainotes/shared/createplan/
# HTML: nginx.ai-servicers.com/{username}/createplan/{descriptor}/