Ftw

FTW (First Try Works)

Arguments: $ARGUMENTS

Parse arguments to determine mode:

Mode Detection

If $ARGUMENTS[0] == "mini":
  MODE = "mini"
  FEATURE_NAME = $ARGUMENTS[1] or null (will ask during discovery)
  PROJECT_PATH = $ARGUMENTS[2] or current working directory

Else if $ARGUMENTS[0] == "init":
  MODE = "init"
  PROJECT_PATH = $ARGUMENTS[1] or current working directory

Else:
  MODE = "full"
  If $ARGUMENTS[0] ends with ".md":
    PRD_PATH = $ARGUMENTS[0]
    PROJECT_PATH = dirname(dirname(PRD_PATH))
    START_PHASE = $ARGUMENTS[1] or 1
    END_PHASE = $ARGUMENTS[2] or auto-detect from PRD
    PRD_NAME = basename without extension
  Else:
    PROJECT_PATH = $ARGUMENTS[0] or current working directory
    START_PHASE = $ARGUMENTS[1] or 1
    END_PHASE = $ARGUMENTS[2] or 4
    PRD_PATH = auto-discover from PROJECT_PATH/PRDs/
    PRD_NAME = discovered PRD basename

Usage

/ftw [prd-path.md] [start] [end]     Full PIV mode with PRD
/ftw [project-path] [start] [end]    Full PIV mode, auto-discover PRD
/ftw                                  Full PIV mode, cwd
/ftw mini [feature-name] [path]       Mini mode (discovery-driven)
/ftw init [project-path]              Project setup

Required Reading by Role

CRITICAL: Each role MUST read their instruction files before acting.

Orchestrator Philosophy

"Context budget: ~15% orchestrator, 100% fresh per subagent"

You are the orchestrator. You stay lean and manage workflow. You DO NOT execute PRPs yourself - you spawn specialized sub-agents with fresh context for each task.

Sub-agent spawning: Use the sessions_spawn tool to create fresh sub-agent sessions. Each spawn is non-blocking - you'll receive results via an announce step. Wait for each agent's results before proceeding to the next step.

Mode: Init

If MODE is "init", set up the project:

mkdir -p PROJECT_PATH/PRDs PROJECT_PATH/PRPs/templates PROJECT_PATH/PRPs/planning

Copy {baseDir}/assets/prp_base.md to PROJECT_PATH/PRPs/templates/prp_base.md if it doesn't exist. Create PROJECT_PATH/WORKFLOW.md from {baseDir}/assets/workflow-template.md if it doesn't exist.

Output: "Project initialized at PROJECT_PATH". Done.

Mode: Full PIV

Prerequisite: A PRD must exist. If none found, tell user to create one first.

For each phase from START_PHASE to END_PHASE, you will execute this pipeline:

1. Check/Generate PRP — Look for existing PRP; if none, spawn a research sub-agent (sessions_spawn) to do codebase analysis + PRP generation
2. Spawn EXECUTOR — Fresh sub-agent (sessions_spawn) implements the PRP requirements
3. Spawn VALIDATOR — Fresh sub-agent (sessions_spawn) independently verifies ALL requirements against the PRP
4. Debug Loop (max 3x) — If GAPS_FOUND: spawn DEBUGGER sub-agent, then re-validate. If HUMAN_NEEDED: ask user
5. Smart Commit — Semantic commit with Built with FTW (First Try Works)
6. Update WORKFLOW.md — Mark phase complete with validation results
7. Next Phase — Loop back to step 1

Read {baseDir}/references/workflow-full.md for the detailed step-by-step instructions including sub-agent prompt templates.

Mode: Mini

No PRD required — starts from a quick conversation with the user.

1. Discovery — Ask 3-5 targeted questions (what it does, where it lives, success criteria, out of scope). Structure answers into YAML
2. Research & PRP Generation — Spawn sub-agent (sessions_spawn) for codebase analysis + PRP generation from discovery answers
3. Spawn EXECUTOR — Fresh sub-agent implements the PRP
4. Spawn VALIDATOR — Fresh sub-agent verifies requirements (or self-validate if <5 files/<100 lines changed)
5. Debug Loop (max 3x) — If GAPS_FOUND: spawn DEBUGGER, re-validate. If HUMAN_NEEDED: ask user
6. Smart Commit — feat(mini): implement {FEATURE_NAME} with Built with FTW (First Try Works)

Read {baseDir}/references/workflow-mini.md for the detailed step-by-step instructions including discovery questions, sub-agent prompt templates, and validation sizing logic.

Error Handling

• No PRD (full mode): Tell user to create one first
• Executor BLOCKED: Ask user for guidance
• Validator HUMAN_NEEDED: Ask user for guidance
• 3 debug cycles exhausted: Escalate with persistent issues list

Sub-Agent Timeout/Failure

When a sub-agent times out or fails:

1. Check for partial work (files created, tests written)
2. Retry once with a simplified, shorter prompt
3. If retry fails, escalate to user with what was accomplished

Quick Reference

File Naming

PRDs/{name}.md                                # PRD (full mode)
PRPs/PRP-{prd-name}-phase-{N}.md              # PRP (full mode)
PRPs/planning/{prd-name}-phase-{N}-analysis.md # Analysis (full mode)
PRPs/mini-{feature-name}.md                   # PRP (mini mode)
PRPs/planning/mini-{feature-name}-analysis.md  # Analysis (mini mode)

Completion

Full Mode

## FTW COMPLETE

Phases Completed: START to END
Total Commits: N
Validation Cycles: M

### Phase Summary:
- Phase 1: [feature] - validated in N cycles
...

All phases successfully implemented and validated.

Mini Mode

## FTW MINI COMPLETE

Feature: {FEATURE_NAME}
Project: {PROJECT_PATH}

### Artifacts
- PRP: PRPs/mini-{FEATURE_NAME}.md
- Analysis: PRPs/planning/mini-{FEATURE_NAME}-analysis.md

### Implementation
- Validation cycles: {N}
- Debug iterations: {M}

### Files Changed
{list}

All requirements verified and passing.