prd.json

{
  "name": "Emergent Learning Framework - Ralph Loop",
  "version": "1.0.0",
  "description": "Autonomous task executor using fresh Claude Code sessions per story",
  "created_at": "2026-01-15",
  "stories": [
    {
      "id": "RALPH-001",
      "title": "Initialize Ralph Loop Infrastructure",
      "description": "Set up the core Ralph Loop system with bash orchestrator, PRD structure, and progress tracking",
      "priority": 1,
      "status": "blocked",
      "acceptance_criteria": [
        "ralph.sh exists and is executable",
        "prd.json follows correct structure",
        "prompt.md is generated correctly for each iteration",
        "progress.txt tracks learnings across sessions",
        "Pre-commit hook calls ralph.sh",
        "Documentation explains the architecture"
      ],
      "files": [
        "tools/scripts/ralph.sh",
        "prd.json",
        "prompt.md",
        "progress.txt",
        "tools/hooks/pre-commit",
        "library/guides/ralph-loop-guide.md"
      ]
    },
    {
      "id": "RALPH-002",
      "title": "Integrate with Claude Code Sessions",
      "description": "Ensure each ralph.sh iteration spawns a fresh claude-code session with proper context",
      "priority": 2,
      "status": "pending",
      "acceptance_criteria": [
        "claude-code invocation reads prompt.md",
        "Session can access git history and current PRD",
        "Session updates progress.txt with learnings",
        "Session commits work automatically",
        "ralph.sh correctly detects session success/failure"
      ],
      "files": [
        "tools/scripts/ralph.sh",
        "prompt.md"
      ]
    },
    {
      "id": "RALPH-003",
      "title": "Create Story Initialization Script",
      "description": "Build init-ralph.sh to help users create their first prd.json from requirements",
      "priority": 3,
      "status": "pending",
      "acceptance_criteria": [
        "init-ralph.sh prompts for project name and requirements",
        "Generates valid prd.json with initial stories",
        "Creates progress.txt with timestamp",
        "Provides quick-start instructions"
      ],
      "files": [
        "tools/scripts/init-ralph.sh",
        "tools/scripts/ralph.sh"
      ]
    },
    {
      "id": "RALPH-004",
      "title": "Test Ralph Loop End-to-End",
      "description": "Run through a complete Ralph Loop cycle with a sample PRD to verify it works",
      "priority": 4,
      "status": "pending",
      "acceptance_criteria": [
        "Sample PRD completes all stories",
        "Each iteration spawns fresh session",
        "progress.txt accumulates learnings",
        "No context degradation detected",
        "Git history shows incremental commits"
      ],
      "files": [
        "tools/scripts/ralph.sh",
        "prd.json",
        "progress.txt"
      ]
    },
    {
      "id": "CARD-001",
      "title": "View ELF Context Card in Ralph Loop",
      "description": "As a developer running a Ralph loop, I want to see a compact JSON card displaying my ELF context (TIER 0, relevant golden rules, current domain) so that I have context at a glance without switching windows",
      "priority": 1,
      "status": "pending",
      "acceptance_criteria": [
        "Card displays TIER 0 project context",
        "Card shows top 5 relevant golden rules for current domain",
        "Card is readable within 1280px viewport",
        "JSON is syntax-highlighted (keys, values, strings colored differently)",
        "Card has 'Collapse all' button to minimize nested objects",
        "Card loads in <500ms",
        "Works offline (no external CDN dependencies)"
      ],
      "files": [
        "src/components/ContextCard.tsx",
        "src/components/ContextCard.css"
      ]
    },
    {
      "id": "CARD-002",
      "title": "Search Context Card by Domain or Keyword",
      "description": "As a developer in a deep project with many heuristics, I want to search the JSON card by domain, confidence level, or keyword so that I can find relevant heuristics without scrolling",
      "priority": 2,
      "status": "pending",
      "acceptance_criteria": [
        "Search is case-insensitive",
        "Highlights matching keys and values",
        "Returns count of matches",
        "Filters work with nested objects",
        "Search input is sticky (stays visible when scrolling)",
        "Supports filters: domain:react, confidence:>0.8"
      ],
      "files": [
        "src/components/ContextCard.tsx",
        "src/components/SearchFilter.tsx"
      ]
    },
    {
      "id": "CARD-003",
      "title": "Export Context JSON for Debugging",
      "description": "As a developer whose Ralph loop failed, I want to export the context JSON that caused the failure so that I can save it, analyze it later, or share it with someone for help",
      "priority": 2,
      "status": "pending",
      "acceptance_criteria": [
        "Export button copies raw JSON to clipboard",
        "Export button downloads JSON file with timestamp",
        "Exported JSON is valid and parseable",
        "Includes metadata: timestamp, domain, Ralph iteration number",
        "File is named predictably: context-TIMESTAMP.json"
      ],
      "files": [
        "src/components/ContextCard.tsx",
        "src/utils/export.ts"
      ]
    },
    {
      "id": "CARD-004",
      "title": "Refresh Context from ELF Building",
      "description": "As a developer running multiple Ralph iterations, I want to refresh the context card to see updated heuristics and progress so that my card stays in sync with the ELF building",
      "priority": 3,
      "status": "pending",
      "acceptance_criteria": [
        "Refresh button fetches latest from ~/.claude/emergent-learning",
        "Shows 'Last updated: X seconds ago' timestamp",
        "Auto-refresh option every 30s (configurable)",
        "Highlights new or changed rules since last load",
        "Gracefully handles offline/unavailable building"
      ],
      "files": [
        "src/components/ContextCard.tsx",
        "src/hooks/useElfContext.ts"
      ]
    },
    {
      "id": "CARD-005",
      "title": "Navigate Nested Context with Breadcrumbs",
      "description": "As a developer exploring deeply nested context (TIER 1 golden rules with multiple levels), I want breadcrumb navigation to jump between levels so that I don't get lost in the structure",
      "priority": 3,
      "status": "pending",
      "acceptance_criteria": [
        "Breadcrumbs show current path (e.g., TIER 1 > golden-rules > [RULE-001])",
        "Click any breadcrumb to jump to that level",
        "Breadcrumbs stay visible while scrolling",
        "Works with any nesting depth"
      ],
      "files": [
        "src/components/ContextCard.tsx",
        "src/components/Breadcrumbs.tsx"
      ]
    }
  ]
}