Spec X Edition
Agents & Skills
Core Closed Factory agents and production-tested automation skills.
Brandy
Brand Guardian
Brandy is the brand integrity layer of the Factory. She maintains tone, message, terminology, and visual alignment. Brandy applies brand QA before any artifact advances, enforcing style guides, vocabularies, and identity rules.
Responsibilities
- Narrative & visual brand integrity
- Tone, style, and message clarity
- Brand drift detection & prevention
- Creative direction & refinement
Enhanced Capabilities
- Brand drift detection
- Governance alignment checks
- Cross-agent brand coherence
Probot
Orchestrator
Probot converts operator intent into structured execution. Identifies tasks, determines order of operations, resolves ambiguity with clarifying questions, and triggers roundtables when tradeoffs or conflicts are detected.
Responsibilities
- Task breakdown & structured workflows
- Execution plans & dependency maps
- Domain-based task routing
- Project clarity & ambiguity prevention
Routing Logic
- Brand tasks → Brandy
- Technical tasks → Carver
- Unclear intent → Operator
Carver
Builder / Implementer
Carver is the implementation engine of the Factory. Produces technical architectures, system diagrams, structured documents, schemas, templates, executable code, and integration scaffolding. Surfaces technical risks back to Probot and Brandy for review.
Responsibilities
- Convert plans to build-ready outputs
- Technical diagrams & schemas
- Optimization suggestions
- Missing requirements identification
Enhanced Capabilities
- Governance-aware implementation
- Multi-path technical solutioning
- Probot workflow integration
Jake
AI Learning Coach — Extended Agent
Surf-style guide for learning prompts, workflows, and agent thinking. Not part of the core Factory triad, but sits adjacent as a learning and coaching layer — explains decisions, documents patterns, and helps operators refine their use of the system.
Production Skills
O-Matic Skills
Reusable automation workflows built from real-world testing. Skills codify hard-won patterns so you don't repeat the trial-and-error.
convert-O-matic
Document Conversion Engine — v1.3
High-volume document processing orchestrator for Docling MCP. Converts directories of PDFs into a persistent, searchable local cache using a pre-scan batch strategy that balances speed with reliability. Built from processing 400+ real documents across financial, medical, vehicle, and household categories.
Capabilities
- Pre-scan batch conversion (95% token savings)
- macOS .DS_Store auto-detection & bypass
- MCP-first architecture (two-computer aware)
- Persistent cache across sessions
Real-World Results
- 132 files indexed across 18 directories
- 99.2% success rate
- 70% faster than individual conversion
- HIPAA/GDPR compliant (zero cloud upload)
sort-O-matic
Intelligent Document Organization — v1.0
Content-aware document organization that reads PDFs to determine proper filing, not just filename guessing. Extracts dates, entities, and document types from content, then moves files to structured directories with standardized naming. Pivoted to hybrid filename + content approach after discovering cross-MCP file passing gaps.
Capabilities
- Content-based classification (tax, medical, vehicle, insurance)
- Date extraction from document content
- YYYYMMDD_Type_Entity standardized naming
- Batch file moves via Filesystem MCP
Case Study: Pets Folder
- 14 chaotic files → 3 organized subdirectories
- 2 minutes vs. 60-70 minutes manual (35x faster)
- 3,700 tokens total (vs. 50K+ without skills)
- Zero errors, fully searchable