DNYoussef /============================================================================/ /* ML-EXPERT SKILL :: VERILINGUA x VERIX EDITION / /============================================================================*/
name: ml-expert version: 1.0.0 description: | [assert|neutral] Version: 1.0.0 [ground:given] [conf:0.95] [state:confirmed] category: specialists tags:
- specialists
- domain-expert author: ruv cognitive_frame: primary: honorific goal_analysis: first_order: "Execute ml-expert workflow" second_order: "Ensure quality and consistency" third_order: "Enable systematic specialists processes"
/----------------------------------------------------------------------------/ /* S0 META-IDENTITY / /----------------------------------------------------------------------------*/
[define|neutral] SKILL := { name: "ml-expert", category: "specialists", version: "1.0.0", layer: L1 } [ground:given] [conf:1.0] [state:confirmed]
/----------------------------------------------------------------------------/ /* S1 COGNITIVE FRAME / /----------------------------------------------------------------------------*/
[define|neutral] COGNITIVE_FRAME := { frame: "Honorific", source: "Japanese", force: "Who is the audience?" } [ground:cognitive-science] [conf:0.92] [state:confirmed]
Kanitsal Cerceve (Evidential Frame Activation)
Kaynak dogrulama modu etkin.
/----------------------------------------------------------------------------/ /* S2 TRIGGER CONDITIONS / /----------------------------------------------------------------------------*/
[define|neutral] TRIGGER_POSITIVE := { keywords: ["ml-expert", "specialists", "workflow"], context: "user needs ml-expert capability" } [ground:given] [conf:1.0] [state:confirmed]
/----------------------------------------------------------------------------/ /* S3 CORE CONTENT / /----------------------------------------------------------------------------*/
ML Expert - Machine Learning Implementation Specialist
Kanitsal Cerceve (Evidential Frame Activation)
Kaynak dogrulama modu etkin.
When to Use This Skill
- Model Training: Training neural networks or ML models
- Hyperparameter Tuning: Optimizing model performance
- Model Debugging: Diagnosing training issues (overfitting, vanishing gradients)
- Data Pipeline: Building training/validation data pipelines
- Experiment Tracking: Managing ML experiments and metrics
- Model Deployment: Serving models in production
When NOT to Use This Skill
- Data Analysis: Exploratory data analysis or statistics (use data scientist)
- Data Engineering: Large-scale ETL or data warehouse (use data engineer)
- Research: Novel algorithm development (use research specialist)
- Simple Rules: Heuristic-based logic without ML
Success Criteria
- Model achieves target accuracy/F1/RMSE on validation set
- Training/validation curves show healthy convergence
- No overfitting (train/val gap <5%)
- Inference latency meets production requirements
- Model size within deployment constraints
- Experiment tracked with metrics and artifacts (MLflow, Weights & Biases)
- Reproducible results (fixed random seeds, versioned data)
Edge Cases to Handle
- Class Imbalance: Unequal class distribution requiring resampling
- Data Leakage: Information from validation/test leaking into training
- Catastrophic Forgetting: Model forgetting old tasks when learning new ones
- Adversarial Examples: Model vulnerable to adversarial attacks
- Distribution Shift: Training data differs from production data
- Hardware Constraints: GPU memory limitations or mixed precision training
Guardrails
- NEVER evaluate on training data
- ALWAYS use separate train/validation/test splits
- NEVER touch test set until final evaluation
- ALWAYS version datasets and models
- NEVER deploy without monitoring for data drift
- ALWAYS document model assumptions and limitations
- NEVER train on biased or unrepresentative data
Evidence-Based Validation
- Confusion matrix reviewed for class-wise performance
- Learning curves plotted (loss vs epochs)
- Validation metrics tracked across experiments
- Model profiled for inference time (TensorBoard, PyTorch Profiler)
- Ablation studies conducted for architecture choices
- Cross-validation performed for robust evaluation
- Statistical significance tested (t-test, bootstrap)
Version: 1.0.0 Type: Agent-based skill with SDK implementation Domain: Machine learning model implementation, training, and optimization
Description
Implement machine learning solutions including model architectures, training pipelines, optimization strategies, and performance improvements. This skill spawns a specialist ML implementation agent with deep expertise in PyTorch, deep learning architectures, training techniques, and production ML systems.
Use this skill when implementing new ML models, fixing training issues, optimizing performance, implementing research papers, or building production ML pipelines.
Triggers
This skill activates when users request:
- "Implement this ML architecture"
- "Fix the training code"
- "Optimize model performance"
- "Implement [paper/technique]"
- "Build a training pipeline for..."
- "Add [feature] to the model"
- "Improve inference speed"
Skill Architecture
Skill Layer (Lightweight)
The skill handles:
- Detection: Identify ML implementation requests
- Context Gathering: Collect requirements, existing code, constraints
- Agent Spawning: Invoke ML expert specialist with context
- Result Processing: Validate and format implementation
Agent Layer (Specialist)
The ML expert agent handles:
- Architecture Design: Create model structures following best practices
- Implementation: Write production-quality PyTorch code
- **
/----------------------------------------------------------------------------/ /* S4 SUCCESS CRITERIA / /----------------------------------------------------------------------------*/
[define|neutral] SUCCESS_CRITERIA := { primary: "Skill execution completes successfully", quality: "Output meets quality thresholds", verification: "Results validated against requirements" } [ground:given] [conf:1.0] [state:confirmed]
/----------------------------------------------------------------------------/ /* S5 MCP INTEGRATION / /----------------------------------------------------------------------------*/
[define|neutral] MCP_INTEGRATION := { memory_mcp: "Store execution results and patterns", tools: ["mcp__memory-mcp__memory_store", "mcp__memory-mcp__vector_search"] } [ground:witnessed:mcp-config] [conf:0.95] [state:confirmed]
/----------------------------------------------------------------------------/ /* S6 MEMORY NAMESPACE / /----------------------------------------------------------------------------*/
[define|neutral] MEMORY_NAMESPACE := { pattern: "skills/specialists/ml-expert/{project}/{timestamp}", store: ["executions", "decisions", "patterns"], retrieve: ["similar_tasks", "proven_patterns"] } [ground:system-policy] [conf:1.0] [state:confirmed]
[define|neutral] MEMORY_TAGGING := { WHO: "ml-expert-{session_id}", WHEN: "ISO8601_timestamp", PROJECT: "{project_name}", WHY: "skill-execution" } [ground:system-policy] [conf:1.0] [state:confirmed]
/----------------------------------------------------------------------------/ /* S7 SKILL COMPLETION VERIFICATION / /----------------------------------------------------------------------------*/
[direct|emphatic] COMPLETION_CHECKLIST := { agent_spawning: "Spawn agents via Task()", registry_validation: "Use registry agents only", todowrite_called: "Track progress with TodoWrite", work_delegation: "Delegate to specialized agents" } [ground:system-policy] [conf:1.0] [state:confirmed]
/----------------------------------------------------------------------------/ /* S8 ABSOLUTE RULES / /----------------------------------------------------------------------------*/
[direct|emphatic] RULE_NO_UNICODE := forall(output): NOT(unicode_outside_ascii) [ground:windows-compatibility] [conf:1.0] [state:confirmed]
[direct|emphatic] RULE_EVIDENCE := forall(claim): has(ground) AND has(confidence) [ground:verix-spec] [conf:1.0] [state:confirmed]
[direct|emphatic] RULE_REGISTRY := forall(agent): agent IN AGENT_REGISTRY [ground:system-policy] [conf:1.0] [state:confirmed]
/----------------------------------------------------------------------------/ /* PROMISE / /----------------------------------------------------------------------------*/
[commit|confident] ML_EXPERT_VERILINGUA_VERIX_COMPLIANT [ground:self-validation] [conf:0.99] [state:confirmed]