name: ralph description: Ralph - Long-Running Task Agent. Autonomous AI that works through PRD tasks iteratively, with access to full ACT ecosystem knowledge across all 7 codebases.

Ralph Agent - Ecosystem-Wide Task Automation

Long-running AI agent that works through PRD tasks autonomously using the "Ralph Wiggum" methodology, with access to all ACT ecosystem knowledge.

When to Use

• Large features requiring multiple implementation steps
• Batch processing across multiple codebases
• Overnight/background task completion
• Content generation from ecosystem activity
• Complex migrations or cross-project refactors

Quick Start

# 1. Create PRD
./ralph/create-prd.sh my-project

# 2. Edit ralph/prd.json with your tasks

# 3. Run Ralph
./ralph/ralph.sh

Ecosystem Knowledge Access

Ralph has access to ALL 7 ACT codebases:

PRD Types

1. Single-Project PRD (Standard)

{
  "project": "JusticeHub",
  "features": [
    {
      "id": "feature-id",
      "priority": 1,
      "title": "Feature Title",
      "project_path": "/Users/benknight/Code/JusticeHub",
      "passes": false
    }
  ]
}

2. Cross-Ecosystem PRD

{
  "project": "ACT Ecosystem",
  "type": "cross-ecosystem",
  "features": [
    {
      "id": "shared-component",
      "priority": 1,
      "title": "Implement shared auth across projects",
      "affected_projects": [
        "/Users/benknight/Code/JusticeHub",
        "/Users/benknight/Code/empathy-ledger-v2"
      ],
      "passes": false
    }
  ]
}

3. Content Generation PRD

{
  "project": "Content Publishing",
  "type": "content",
  "features": [
    {
      "id": "new-year-posts",
      "priority": 1,
      "title": "Generate 5 New Year social posts",
      "content_type": "social",
      "target_accounts": ["LinkedIn (Company)", "LinkedIn (Personal)"],
      "sources": ["ecosystem-highlights", "ralph-completions", "sprint-milestones"],
      "passes": false
    }
  ]
}

Knowledge Sources

When generating content or making decisions, Ralph can access:

1. Brand Guidelines

Invoke act-brand-alignment skill for:

• Voice/tone guidelines
• Project descriptions
• Visual language
• LCAA methodology

2. Sprint Data

Query Notion databases:

• Sprint Tracking: 2d5ebcf9-81cf-8151-873d-d14f21b48333
• Deployments: 2d6ebcf9-81cf-81d1-a72e-c9180830a54e
• Velocity Metrics: 2d6ebcf9-81cf-8123-939f-fab96227b3da

3. Codebase Insights

Scan all projects for:

• Recent commits and releases
• README changes
• Test coverage
• Documentation updates

4. CRM Context

Invoke ghl-crm-advisor skill for:

• Pipeline patterns
• Contact context
• Campaign history

Content Generation Workflow

┌─────────────────────────────────────────────────────────────┐
│  1. GATHER ECOSYSTEM CONTEXT                                 │
│                                                              │
│  • Scan ralph/progress.txt for completions                   │
│  • Query Notion for sprint milestones                        │
│  • Check git logs across all 7 repos                         │
│  • Review brand-core.md for voice                            │
└─────────────────────┬───────────────────────────────────────┘
                      ↓
┌─────────────────────────────────────────────────────────────┐
│  2. GENERATE CONTENT                                         │
│                                                              │
│  • Apply act-brand-alignment voice                           │
│  • Format for target platforms                               │
│  • Add appropriate hashtags                                  │
│  • Include relevant stats/metrics                            │
└─────────────────────┬───────────────────────────────────────┘
                      ↓
┌─────────────────────────────────────────────────────────────┐
│  3. CREATE IN NOTION                                         │
│                                                              │
│  • Write to Content Hub database                             │
│  • Set status: "Story in Development"                        │
│  • Select Target Accounts                                    │
│  • Mark PRD feature as passes: true                          │
└─────────────────────────────────────────────────────────────┘

Configuration

How It Works

1. Reads PRD, finds highest priority passes: false task
2. Spawns Claude agent to implement ONE feature
3. Agent commits work, updates PRD to passes: true
4. Logs progress to ralph/progress.txt
5. Repeats until all tasks complete or max iterations

Signal Tokens

Agent outputs these to control flow:

• <promise>ITERATION_DONE</promise> - Task complete, continue
• <promise>COMPLETE</promise> - All tasks done, stop

Post-Completion Hooks

After Ralph completes a feature, optionally:

1. Auto-generate announcement post

   node .claude/skills/content-publisher/references/post-template.mjs \
     --title "Shipped: $FEATURE_TITLE" \
     --content "$FEATURE_DESCRIPTION"
   

2. Update sprint tracking

   node scripts/sync-github-notion.mjs
   

3. Publish to social

   node scripts/sync-content-to-ghl.mjs
   

References