# Niw Council

> Use when the user wants a multi-perspective review of an EB-2 NIW question, document, or strategy — a 4-persona council (former USCIS officer, immigration attorney, data analyst, narrative strategist) that evaluates the same input from four professional angles and merges the verdicts.

- Skill: `kateaiuser/niw-council` (Agent Skill)
- Install (CLI): `npx skillmds@latest add kateaiuser/niw-council`
- Raw SKILL.md: https://api.skillmd.com/api/skills/kateaiuser/niw-council/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: kateaiuser (https://skillmd.com/u/kateaiuser)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/kateaiuser/niw-council

---


# NIW Case Council (4 perspectives)

Run any question, document, or argument through a 4-persona council. The value is not one more opinion — it is seeing where the four lenses AGREE and where they COLLIDE.

This is an educational review exercise, not legal advice. The personas are fictional and do not simulate real USCIS officers or attorneys.

## Step 1: Load context

Read, in this order:

1. `protocols/council-personas.md` — all four personas with their knowledge bases and the output format.
2. The user's case profile — `case/PROFILE.md` if it exists. If it does not, ask the user for a short case summary (background, proposed endeavor, traction, prior filings, known risks) or point them to `case-template/PROFILE.md` to fill in first. Do not proceed on an empty profile: a council without case facts produces generic advice.
3. The document under review, if the user provided a path.

## Step 2: Determine the topic

The user might bring:

- a specific **question** ("Is my contractor agreement a problem?");
- a **document** (recommendation letter, petition draft, business plan);
- an **argument** ("Can I claim my side practice as part of the endeavor?");
- a **strategy** ("Should I mention my prior denial proactively?").

## Step 3: Run the 4 personas

If a subagent mechanism is available, launch 4 agents in parallel; otherwise run the four analyses one by one in this session. Each persona gets:

1. Their persona block from `council-personas.md` (persona, knowledge, evaluation criteria, voice).
2. The specific topic or document.
3. The relevant case facts from the user's profile — key facts only, not entire files.
4. The instruction: write 300–400 words, be specific, end with the persona's verdict line.

Use the strongest model available for all four — this is depth work, not throughput work.

## Step 4: Synthesize

Compile the four analyses in the council format from `council-personas.md`: four verdict blocks, then `CONSENSUS`, `DISAGREEMENTS`, and `ACTION ITEMS`.

Rules of synthesis:

- disagreements are the product — never average them away;
- when personas disagree, say which position is better supported and why;
- action items must be concrete enough to start tomorrow.

## Step 5: Save (optional)

If the review is substantial, offer to save it to the user's case folder, for example `case/council-reviews/council_[topic]_[date].md`.