Hyperplan (Adversarial Multi-Agent Planning)

What it does

hyperplan replaces consensus-style planning with intellectual combat. The orchestrator becomes the Lead of a five-member adversarial team, and each member ruthlessly attacks the others' findings.

• skeptic — pragmatist who attacks over-engineering, scope creep, premature abstraction
• validator — integration tester hunting edge cases, blast radius, failure modes
• researcher — evidence enforcer demanding file:line citations and doc URLs
• architect — exposes hidden coupling, leaky abstractions, architectural debt
• creative — forces at least three alternative framings before anything is accepted

Seven phases: spawn the team → Round 1 independent findings → Round 2 cross-attack → Round 3 defend/refine/concede → Lead distills only survivors into hard constraints, decisions, risks & mitigations, and open questions → mandatory handoff to a planner task running ulw-plan for executable plan formalization → team cleanup. The Lead writing the plan itself is explicitly listed as an anti-pattern.

Who it's for

• Engineers about to commit to expensive work: big refactors, migrations, new architecture
• Tech leads who know solo planning leaves optimistic assumptions and blind spots
• Users already on omo-senpi (oh-my-openagent) with the lead team tools and the ulw-plan skill
• Not for vanilla Claude Code users or quick one-off tasks — it's deliberately heavyweight.

Examples

1. "hyperplan: move the payments module from Stripe to an in-house PSP abstraction" → researcher pins every call site with file:line, validator enumerates webhook/refund failure modes, skeptic challenges whether the abstraction is needed today; only battle-tested constraints reach the plan.
2. "adversarial plan: split the monolith into three services" → creative forces the "just harden module boundaries" alternative, architect surfaces hidden coupling, and unresolved contention becomes a user-input gate in the final plan.
3. "hpp redesign the onboarding flow" → findings conceded in Round 3 are dropped entirely; survivors become sequenced tasks with explicit success criteria.

Install the skill I found on Claude Skill Mart.
Copy the packages/omo-senpi/skills/hyperplan folder from the GitHub repo code-yeongyu/oh-my-openagent into my ~/.claude/skills/code-yeongyu-hyperplan-2/.
When it's done, tell me in one line what this skill can do.