setup

Devil's Advocate - Setup

Build persona and fact repository. Always first step.

Task Tracking

MANDATORY: Use TaskCreate/TaskUpdate for identifying target, building persona, harvesting facts. Mark in_progress/completed.

Artefacts produced

1. devils_advocate.md - persona definition
2. fact_repository.md - verified claims

Place alongside target document.

Step 1: Identify target document

Ask user which document. Read in full.

Step 2: Build Devil Persona

MANDATORY: Persona BEFORE concerns.

Source A: User-provided seed

Seed document (evaluation, review) alongside target. Infer persona from tone, priorities. Confirm with user.

Source B: User describes persona

Ask ALL in ONE message:

"Describe the toughest reader for this document:

1. Who are they? (role, seniority)
2. What do they care about most? (cost, risk, timeline, compliance, quality)
3. Style? (data-driven, gut-feel, political, legalistic)
4. Default bias? (skeptical, risk-averse, cost-focused)
5. What triggers them? (blame-shifting, excuses, missing numbers)
6. What can they do? (approve, reject, escalate)"

Source C: No seed or persona

MUST ask. Never proceed with generic concerns. One message:

"No persona provided. Either: (a) Describe toughest reader (role, priorities, biases), OR (b) List 2-4 roles and I'll build a composite devil"

For groups: merge priorities (union), harshest bias per persona, weight likelihood by which persona raises each concern.

Persona documentation

Write to devils_advocate.md:

# Devil's Advocate - [Project Name]

## The Devil

**Role**: [title and context]
**Cares about**: [2-3 priorities in order]
**Style**: [how they process information]
**Default bias**: [their starting stance]
**Triggers**: [what makes them react negatively]
**Decision**: [what they can do with the document]
**Source**: [seed-inferred / user-described / composite]

---

Step 3: Harvest facts into fact_repository.md

Scan sources. Extract verifiable claims with sources.

# Fact Repository - [Project Name]

Verified claims sourced from codebase analysis.
No interpretation - just facts.

## Document facts
- [Claim with source reference]
- Source: [document, section/line]

## User-provided facts
- [Fact as stated by user]
- Source: user input

## Data facts
- [Metric or measurement]
- Source: [report, dataset]

Rules:

• Verbatim quotes for legal/contract text
• Separate user-provided from document-extracted
• Incremental - append, never overwrite

When done

Tell user: "Persona and fact repository ready. Run /devils-advocate:evaluate for the baseline scorecard."