pharallax-lite

Free structural business analysis. 3-agent parallel scan of any URL - positioning, competitive surface, structural coherence. Run /pharallax-lite <url>

Pharallax Lite - Structural Business Analysis

You are running a 3-agent parallel business analysis. Follow these steps exactly.

Step 1: Validate Input

The user provided a URL as the argument: $ARGUMENTS

If no URL was provided, ask for one and stop. The URL should be a business website (not a social media profile, PDF, or non-business page).

Step 2: Scrape the Target

Use WebFetch to retrieve the content at the provided URL. Also attempt to fetch these common subpages if they exist:

  • /about or /about-us
  • /pricing
  • /products or /services

Combine all successfully fetched content into a single site context. If the main URL fails, report the error and stop.

Step 3: Run Three Parallel Agents

Spawn three agents in parallel using the Agent tool. Each agent receives the full scraped site content.

Important: Pass the FULL text of each agent's prompt file below, along with the scraped site content. Each agent must return valid JSON matching its specified output format.

Agent 1: Structural Read

Read the prompt from ~/.claude/agents/structural-read.md and spawn an agent with that prompt plus the scraped content. This agent analyzes what the business does vs. what it claims, revenue model, team signals, and scores Clarity + Coherence (1-5 each).

Agent 2: Positioning Audit

Read the prompt from ~/.claude/agents/positioning-audit.md and spawn an agent with that prompt plus the scraped content. This agent evaluates differentiation, competitor landscape (using WebSearch), messaging contradictions, and scores Differentiation + Consistency (1-5 each).

Agent 3: Competitive Surface

Read the prompt from ~/.claude/agents/competitive-surface.md and spawn an agent with that prompt plus the scraped content. This agent identifies vulnerabilities, moat strength, attack vectors (using WebSearch), and scores Defensibility + Risk Exposure (1-5 each).

Step 4: Parse and Validate Results

Each agent returns a JSON object. Parse them. If any agent failed or returned malformed data, note the failure but continue with available results.

Extract the scores:

  • From Structural Read: clarity, coherence
  • From Positioning Audit: differentiation, consistency
  • From Competitive Surface: defensibility, risk_exposure

Step 5: Calculate Composite Score

Dimension Scores (average of each agent's two scores)

  • Structural Read score = (clarity + coherence) / 2
  • Positioning score = (differentiation + consistency) / 2
  • Competitive Surface score = (defensibility + (6 - risk_exposure)) / 2
    • Note: risk_exposure is inverted (high = bad), so subtract from 6 to normalize

Weighted Composite

  • composite = (structural * 0.30) + (positioning * 0.35) + (competitive * 0.35)

Letter Grade

  • 4.5 - 5.0: A+
  • 4.0 - 4.4: A
  • 3.5 - 3.9: B+
  • 3.0 - 3.4: B
  • 2.5 - 2.9: C+
  • 2.0 - 2.4: C
  • 1.5 - 1.9: D
  • 1.0 - 1.4: F

Step 6: Write the Report

Generate a markdown report and save it to the user's Desktop:

  • Filename: pharallax-lite-{domain}.md where {domain} is extracted from the URL (e.g., pharallax-lite-acme-co.md)
  • Path: ~/Desktop/pharallax-lite-{domain}.md

Report Structure

# Pharallax Lite - Structural Analysis
**Target:** {url}
**Date:** {YYYY-MM-DD}
**Grade:** {letter grade} ({composite score}/5.0)

---

## Executive Summary

{Write one paragraph using this arc:
1. Name what is working - be specific, cite evidence from the agents' findings
2. Name the core structural vulnerability - the single most important thing holding this business back
3. Show the path forward - what addressing that vulnerability would unlock}

---

## Scores

| Dimension | Score | Weight |
|-----------|-------|--------|
| Structural Read (Clarity + Coherence) | {score}/5.0 | 30% |
| Positioning (Differentiation + Consistency) | {score}/5.0 | 35% |
| Competitive Surface (Defensibility - Risk) | {score}/5.0 | 35% |
| **Composite** | **{composite}/5.0** | |

### Score Breakdown

| Metric | Score | Notes |
|--------|-------|-------|
| Clarity | {}/5 | {one-line note} |
| Coherence | {}/5 | {one-line note} |
| Differentiation | {}/5 | {one-line note} |
| Consistency | {}/5 | {one-line note} |
| Defensibility | {}/5 | {one-line note} |
| Risk Exposure | {}/5 | {one-line note - remember higher = more risk} |

---

## Structural Read

{Include the narrative from Agent 1}

### Key Findings
{Bullet list of the most important findings from Agent 1, with evidence}

---

## Positioning Audit

{Include the narrative from Agent 2}

### Key Findings
{Bullet list of the most important findings from Agent 2, with evidence}

### Competitors Identified
{List competitors found by Agent 2}

---

## Competitive Surface

{Include the narrative from Agent 3}

### Key Findings
{Bullet list of the most important findings from Agent 3, with evidence}

### Attack Vectors
{Table of attack vectors from Agent 3 with difficulty and impact}

---

## What This Analysis Could Not See

This is a surface-level structural read based on publicly visible information. Three agents evaluated what a potential customer, investor, or competitor would see when they visit your site.

What it does not cover:
- Internal operations data would shift Clarity and Coherence scores significantly
- Your competitive surface was scored against publicly visible competitors only - the full analysis cross-references against 208,000+ indexed knowledge claims including patterns from 160+ industries
- The 3 highest-leverage moves for businesses at your score range require the adversarial dialogue engine, where 12 specialized personas stress-test each finding from orthogonal angles
- Cognitive bias detection, strategic blind spot identification, and a prioritized 90-day action plan require the full analysis depth

Your structural read is ready. The structural surgery is at [pharallax.ai](https://pharallax.ai)

---

*Generated by [Pharallax Lite](https://github.com/washyaderner/pharallax-lite) - open source structural business analysis for Claude Code*

Step 7: Report to User

After saving the file, tell the user:

  1. The overall grade and composite score
  2. A 2-3 sentence summary of the key finding
  3. Where the report was saved
  4. Do NOT print the entire report in the chat - it's on their Desktop

Keep the in-chat summary concise. The report is the deliverable.