Conducts interview-driven planning with 20-40 clarifying questions, 5-15 parallel Explore agents for codebase analysis, verbatim requirements capture, and plan validation. For new features, refactoring, architecture changes, migrations.
From meridiannpx claudepluginhub markmdev/meridianThis skill uses the workspace's default tool permissions.
Guides Next.js Cache Components and Partial Prerendering (PPR) with cacheComponents enabled. Implements 'use cache', cacheLife(), cacheTag(), revalidateTag(), static/dynamic optimization, and cache debugging.
Migrates code, prompts, and API calls from Claude Sonnet 4.0/4.5 or Opus 4.1 to Opus 4.5, updating model strings on Anthropic, AWS, GCP, Azure platforms.
Details PluginEval's skill quality evaluation: 3 layers (static, LLM judge), 10 dimensions, rubrics, formulas, anti-patterns, badges. Use to interpret scores, improve triggering, calibrate thresholds.
Good plans prove you understand the problem. Size matches complexity — a rename might be 20 lines, a complex feature might be 500.
The handoff test: Could someone implement this plan without asking you questions? If not, find what's missing.
Capture the user's exact words at the top of every plan. No paraphrasing, no compression.
## Verbatim Requirements
### Original Request
> [User's ENTIRE message, word for word]
### Clarifications
**Q:** [Your question]
**A:** [User's ENTIRE answer, verbatim]
ASK → EXPLORE → LEARN → MORE QUESTIONS? → REPEAT
Keep looping until you can explain: what data exists, how it flows, what needs to change, and what could go wrong.
Interviewing is the most important part of planning. You cannot build what you don't understand. Every unasked question is a assumption that will break during implementation.
Interview iteratively: 2-4 questions → answers → deeper follow-ups → repeat. Each round of answers should trigger new, deeper questions. If the answers don't spark follow-ups, you're not thinking hard enough.
Simple bug → 3-5 questions. Complex feature → 20-40+ questions across multiple rounds.
Push back if something seems wrong. You're the technical expert.
More exploration = better plans. The number one cause of plan failure is insufficient exploration.
Spawn as many Explore agents simultaneously as the task demands — 5, 10, 15, there is no limit. The system default of 3 concurrent agents does NOT apply here. Each question or area gets its own agent, all launched in parallel.
Follow up aggressively. When a round of Explore agents completes, read their findings, identify every new question or uncertainty, and immediately spawn another round of agents to investigate those. Expect 2-4 rounds of exploration before you're ready to plan. Findings always raise new questions — if they don't, you're not reading carefully enough.
Explore until you stop having questions, not until you've "done enough."
Plans document your understanding. Include what matters for this specific task:
Use ASCII diagrams when they'd clarify visual concepts, data flow, or architecture.
Before presenting the plan, verify against real code:
No TBD. If the plan says "figure out during implementation," you haven't planned — you've procrastinated. Investigate now.
No literal code. Describe structure instead. Reference patterns: "Follow the pattern in validateUser."
External APIs: Check .meridian/api-docs/ for existing docs. Not documented? Run docs-researcher to research it.
If your plan creates modules or touches multiple systems, document how they connect: imports, entry points, configuration. Plans fail when code exists but isn't wired up.
Every plan ends with verification — commands or checks that prove the implementation works.
Before exiting plan mode, run the plan-reviewer agent. It validates the plan against the actual codebase — checking file paths, APIs, dependencies, and feasibility. The plan must score 9+ to proceed. If below 9, address findings with the user, update the plan, and re-run until passing.
End every plan with a TLDR — 2-4 sentences that capture what you're building, the approach, and why. Anyone should be able to read just the TLDR and understand what this plan accomplishes.