multi-perspective-scan

Multi-Perspective Scan — The Five Lenses, Standalone

When you ask one question, you get the majority view — one framing, the surface. The breakthrough behind Stanford's STORM is simple: ask the same topic from five independent expert lenses and you catch the blind spots a single pass never sees. This skill runs just that fan-out — fast, when you don't need the full contradiction map / synthesis / peer-review pipeline (for that, use storm-research).

The five lenses, and what each uniquely catches:

Lens	Sees what others miss
Practitioner	The theory–practice gap; what actually works in the field
Skeptic	Overclaims, hidden flaws, the buried failure cases
Economist	Who profits, who pays, the misaligned incentives
Historian	The pattern that repeated before; where we are in the cycle
Academic	What the evidence actually shows, including conflicting findings

Workflow

Step 0 — Scope

1. Get the topic. If vague or sprawling, ask one sharpening question. A tight topic yields sharper perspectives.
2. Derive a slug and set the workspace .storm/{slug}/. Create it: mkdir -p .storm/{slug}.
3. Tune the panel if a swap clearly helps (e.g. add a Clinician / Regulator / End User via a Task call using the same persona output contract). Default to the standard five.

Step 1 — Fan Out (parallel)

Launch all five persona agents in a single message with five Task calls. Each gets the topic, the workspace path, and its output file. Per-agent prompt:

"Topic: {topic}. Workspace: .storm/{slug}/. Write your analysis to .storm/{slug}/{persona}.md per your output contract. Gather real evidence with WebSearch/WebFetch and cite it."

Targets: practitioner.md, skeptic.md, economist.md, historian.md, academic.md. Each agent writes its file and returns a one-line confirmation. If one fails, note it and continue with the rest.

Step 2 — Present the Five Views