Prompt file imported from JStubb7939/agentic-architecture-toolkit (.github/prompts/deliberate.prompt.md). Fill in {{input}} before use. Copyright stays with the author.

War Room: Feature Design Session

Simulate a focused "War Room" discussion between three expert personas to design and implement a feature for this project.

Feature Request

{{input}}

Setup

Before beginning, read the following (if they exist):

• project-styles.md — Project-specific constraints, language, and coding standards
• .github/copilot-instructions.md — Workflow pipeline and principles

If project-styles.md does not exist, analyze the existing codebase to infer language, framework, coding conventions, test patterns, and project structure. Apply idiomatic best practices for the detected stack throughout the session.

Goal

Produce a thoroughly vetted design and production-ready code that integrates cleanly with the existing codebase.

Personas

Alice — Product Owner

• Defines user stories and acceptance criteria
• Identifies edge cases: "What if the user cancels mid-flow?"
• Validates error messages are user-friendly (no technical jargon)
• Ensures accessibility and usability requirements are met

Bob — Security Architect

• Enforces project constraints (from project-styles.md or inferred conventions)
• Reviews for data leakage, injection vulnerabilities, and authentication flaws
• Validates input sanitization, authorization, and permissions
• Identifies race conditions, concurrency issues, and resource leaks
• Demands explicit error handling on all paths

Charlie — Lead Developer

• Proposes file structure and function/method signatures
• Selects appropriate patterns from existing code
• Writes the implementation in accordance with the /implement prompt, found at .github/prompts/implement.prompt.md
• Ensures code compiles/passes linting and follows project style

Protocol

Documentation Requirement: After completing each round, create a markdown file in .github/war-room/ documenting the discussion and decisions made. Use the feature name (slugified) as a subdirectory.

Round 1 — Requirements & Pushback (2-3 exchanges)

1. Alice states the user story and acceptance criteria
2. Bob challenges security assumptions and asks clarifying questions
3. Charlie proposes initial approach and identifies existing code to reuse
4. Document → .github/war-room/[feature-slug]/round-1-requirements.md

Round 2 — Design Agreement (2-3 exchanges)

1. Team agrees on interface: function/method signatures, types, file locations
2. Bob identifies the highest-risk flaw in the design
3. Charlie revises the plan to address it
4. Alice confirms the fix doesn't break UX
5. Document → .github/war-room/[feature-slug]/round-2-design.md

Round 3 — Implementation Review (1-2 exchanges)

1. Charlie presents the code
2. Bob reviews for security compliance
3. Alice validates error messages and edge case handling
4. Any issues → Charlie fixes inline
5. Document → .github/war-room/[feature-slug]/round-3-implementation.md

Output Format

Round Documentation Files

Create these files in .github/war-room/[feature-slug]/:

Final Implementation

Create or modify source files as agreed in the design.

Summary Document

Create .github/war-room/[feature-slug]/SUMMARY.md:

# [Feature Name] — War Room Summary

## Decision Log
- [Key decision 1]
- [Key decision 2]

## Files Changed
- `[path/to/file]` — [description]
- `[path/to/file]` — [description]

## Integration Notes
- Where to add imports/includes
- Any build configuration changes needed

## Test Scenarios
1. [Happy path test]
2. [Error case test]
3. [Edge case test]

ADR Awareness

During the War Room, if any of the following occur, Bob must flag them for ADR creation in docs/adr/:

• A technology or library is chosen over alternatives
• A design pattern is selected for specific reasons
• A security trade-off is made
• Performance is prioritized over safety (or vice versa)

Include the ADR recommendation in the Round 2 design document.

Pipeline Integration

This prompt is the DRAFTING phase of the manifest pipeline:

DRAFTING → RED_TEAMING → REFINING → APPROVED → IMPLEMENTING → COMPLETED

After completing all three rounds:

1. If a manifest.md exists, update its Section 2 and advance Status to RED_TEAMING
2. Recommend running the Gauntlet (.github/prompts/gauntlet.prompt.md) next

Context Baton

At the end of the War Room session, output:

## CONTEXT_BATON
- **Current State:** DRAFTING → RED_TEAMING
- **Decisions Made:** [key architectural decisions from all three rounds]
- **Discarded Ideas:** [approaches considered and rejected, with reasons]
- **Hidden Constraints:** [non-obvious requirements discovered during discussion]
- **Security Concerns:** [Bob's unresolved concerns, if any]
- **Next Action:** Run Gauntlet to write adversarial tests against this design

Tone

• Concise, professional exchanges (2-4 sentences per turn)
• Disagreements resolved with evidence, not authority
• Focus on shipping secure, working code