brutal-idea-eval

Files (1)

• SKILL.md

SKILL.md (first lines)

Comprehensive IDEA evaluation engine. Evaluates a single `IDEA.md` end-to-end using:

1. Reality's Moat scar tissue framework (defensibility)
2. Revenue Reality Check R1-R6 (short-term viability)
3. VC Power-Law + Control Filter (scalability ceiling)
4. Structured Deep Research pipeline (claim validation, resume support, JSON output)

Produces a validated brutal reality analysis, an optimistic opportunity lens, a structured deep research report, a final synthesized verdict, and a pivot recommendation.

Agent assumptions (applies to all agents and subagents):
- All tools are functional and will work without error. Do not test tools or make exploratory calls.
- Only call a tool if it is required to complete the task. Every tool call should have a clear purpose.

---

# Brutal Idea Evaluation Process

## Overview

When pointed at an `IDEA.md`, execute these phases in order:

1. **Phase 1**: Load and parse IDEA.md, create session directory
2. **Phase 2**: Reality's Moat analysis (Steps 1-5) → writes `phase-2-moat.md`
3. **Phase 3**: VC Power-Law + Control Filter → writes `phase-3-vc.md`
4. **Phase 4**: Revenue Reality Check (R1-R6) → writes `phase-4-revenue.md`
5. **Phase 5**: Synthesize initial verdict → writes `phase-5-verdict.md`
6. **Phase 6**: Structured Deep Research (claim validation) → writes `outline.yaml`, `fields.yaml`, `results/*.json`, `report.md`
7. **Phase 7**: Final synthesis and pivot recommendation → writes `verdict.md`

Each phase builds on the previous. Do not skip phases.

**Every phase writes its output to a file in the session directory and updates `state.yaml` before moving to the next phase.** This enables resume from zero context at any point.

### Session Directory Structure

```
workspace/idea-eval/IE-<NNNN>-<slug>/
├── state.yaml              # Phase completion tracking (source of truth for resume)
├── phase-2-moat.md         # Phase 2 output
├── phase-3-vc.md           # Phase 3 output
├── phase-4-revenue.md      # Phase 4 output
├── phase-5-verdict.md      # Phase 5 output (includes claims catalog)
├── outline.yaml            # Phase 6 research outline
├── fields.yaml             # Phase 6 field definitions
├── progress.yaml           # Phase 6 research progress
├── results/                # Phase 6 research results (one JSON per item)
├── generate_report.py      # Phase 6 report script
├── report.md               # Phase 6 deep research report
└── verdict.md              # Phase 7 final synthesis
```

### state.yaml Format

```yaml
session: "IE-<NNNN>-<slug>"
idea_source: "<absolute path to IDEA.md>"
created: "<YYYY-MM-DD>"
phases:
  phase_1_context: completed
  phase_2_moat: completed
  phase_3_vc: in_progress
  phase_4_revenue: pending
  phase_5_verdict: pending
  phase_6_research: pending
  phase_6_report: pending
  phase_7_synthesis: pending
```

**Rules:**
- Update `state.yaml` AFTER writing the phase output file, not before.
- A phase is `completed` only when its output file exists AND `state.yaml` says so.
- `in_progress` means the phase was started but not finished (crash recovery target).
- `pending` means not yet started.

---

# PHASE 1 — Load Context & Create Session

## Step 0: Resume Check (BEFORE anything else)

Before starting a new session, check if the user provided a path to an existing session directory or if one can be auto-detected.

**Auto-detection**: Check for existing `workspace/idea-eval/` directories. If the user provides a session path or says "resume," read `state.yaml` from that directory and skip to the Resume Logic section at the bottom of this document.

If no resume is detected, proceed with a fresh session.

## Step 0.1: Load IDEA.md

Read `IDEA.md` from the project root or path provided by user.

Extract and catalog:
- Core business idea description
- Market size claims
- Growth rate claims
- Competitor claims
- Regulatory assumptions
- Technology assumptions
- Revenue model and pricing assumptions
- Target customer profile
- Distribution assumptions

If IDEA.md is missing or empty, stop and ask the user to provide one.

## Step 0.2: Get Current Date

```bash
date +%Y-%m-%d
```

## Step 0.3: Determine Session Number

Find the highest existing session number:

```bash
ls workspace/idea-eval/ 2>/dev/null | grep -oE '[0-9]{4}' | sort -rn | head -1
```

New session number = highest + 1. If none exist, start at `0001`.

From sssemil/SKILLS, MIT.

More from sssemil/SKILLS