all publishers

ndpvt-web

@ndpvt-web source repo

659 published skills · page 1 of 7

  1. ▌
    Dag · ndpvt-web bundle
    Persistent axiom registry + formal proof prompt builder. Transforms any project into a DAG-structured knowledge base where every conclusion cites parent axioms and inference rules. Axioms co-built with user, saved to .dag/ files, referenced by ID (A1, T1, D1, H1) to save tokens and prevent drift. Use when user says "/dag", "/dag init", "/dag prove", "/dag derive", "formal proof mode", or wants traceable auditable reasoning across sessions. All prose output uses caveman compression.
    14 repo stars
  2. ▌
    Deeptutor Claude Skill · ndpvt-web bundle
    DeepTutor - AI-Powered Personalized Tutoring System
    14 repo stars
  3. ▌
    Swe Pruner Self Adaptive Context Pruning · ndpvt-web
    Apply SWE-Pruner's goal-conditioned context pruning to reduce token usage when working with large codebases. Teaches Claude to selectively skim code like a human programmer: formulate an explicit focus goal, score lines by relevance, and drop low-value content while preserving syntactic structure. Trigger phrases: - "prune the context" / "reduce context size" - "too much code, focus on what matters" - "compress this code for the task" - "skim this file for [specific concern]" - "only show me the relevant parts" - "reduce token usage on this codebase"
    14 repo stars
  4. ▌
    Taming Scylla Understanding Multi Headed · ndpvt-web
    Evaluate and optimize agentic coding tool configurations using Scylla's tiered ablation framework and Cost-of-Pass (CoP) metric. Helps decide what level of agent complexity (prompts, skills, tools, multi-agent) a task actually needs, avoiding overengineering. Trigger phrases: "benchmark my agent setup", "evaluate agent complexity", "Cost-of-Pass analysis", "ablation study for agent config", "optimize agent architecture", "is multi-agent overkill for this task"
    14 repo stars
  5. ▌
    Tricky2 Benchmark Evaluating Human Error · ndpvt-web
    Taxonomy-guided analysis of mixed human+LLM bugs in code. Classifies bug origins, localizes interacting defects, and repairs hybrid-origin errors. Use when: 'review this AI-generated code for bugs', 'find bugs in this human+AI codebase', 'classify whether this bug is human or LLM', 'audit code written by both humans and copilot', 'debug interacting errors in mixed-origin code', 'analyze bug patterns in AI-assisted development'.
    14 repo stars
  6. ▌
    Unikie Bench Benchmarking Multimodal Key · ndpvt-web
    Extract structured key information from document images using schema-guided prompting for LMMs. Builds KIE pipelines that output normalized JSON from receipts, invoices, forms, medical records, tax documents, and more. Trigger phrases: 'extract fields from this document image', 'build a document KIE pipeline', 'parse invoice fields into JSON', 'extract key information from scanned documents', 'structured data extraction from document photos', 'benchmark document extraction accuracy'
    14 repo stars
  7. ▌
    Vista Scene Aware Optimization Streaming · ndpvt-web
    Implement Vista-style scene-aware streaming video processing pipelines with dynamic segmentation, hierarchical compression, and selective recall. Use when building video QA systems, streaming video analysis, real-time surveillance analytics, or any application that processes continuous video and must answer queries against an ever-growing visual context. Trigger phrases: - "streaming video question answering" - "scene-aware video segmentation and compression" - "build a video stream processing pipeline with memory management" - "real-time video QA with GPU/CPU memory offloading" - "long-form video understanding with selective recall" - "process continuous video frames and answer queries efficiently"
    14 repo stars
  8. ▌
    A RAG Scaling Agentic Retrieval Augmented · ndpvt-web
    Build agentic RAG systems where the LLM autonomously decides retrieval strategy using hierarchical interfaces (keyword search, semantic search, chunk read) instead of fixed retrieval pipelines. Replaces single-shot retrieval and predefined workflows with a ReAct-style loop that scales with model capability. Trigger phrases: - "build an agentic RAG system" - "implement hierarchical retrieval" - "make the LLM control its own retrieval" - "scale RAG with reasoning" - "multi-hop question answering over documents" - "replace my fixed RAG pipeline with an agent"
    14 repo stars
  9. ▌
    Agenticpay Multi Agent Negotiation System · ndpvt-web
    Build multi-agent LLM negotiation systems where buyer and seller agents reach deals through natural language. Use when asked to 'build a negotiation system', 'create buyer-seller agents', 'implement price bargaining with LLMs', 'simulate a marketplace with AI agents', 'design an agentic commerce pipeline', or 'build a multi-round trading framework'.
    14 repo stars
  10. ▌
    Alienlm Alienization API Boundary Privacy · ndpvt-web
    Implement AlienLM-style API-boundary privacy layers that protect sensitive text sent to black-box LLM APIs using vocabulary-scale bijective token remapping. Use when: 'add privacy layer to LLM API calls', 'protect prompts sent to external API', 'alienize text for API privacy', 'build token-level encryption for LLM pipeline', 'implement bijective vocabulary mapping', 'privacy-preserving LLM deployment'.
    14 repo stars
  11. ▌
    Amem4rec Leveraging Cross User Similarity · ndpvt-web
    Build agentic recommendation systems that learn collaborative filtering signals through cross-user memory evolution -- no CF model pre-training needed. Use when: 'build a recommender with memory', 'add collaborative filtering to LLM recommendations', 'cross-user pattern memory pool', 'agentic recommender system', 'memory-augmented ranking', 'evolving user behavior patterns for recommendations'.
    14 repo stars
  12. ▌
    Arkeval Benchmarking Evaluating Automated · ndpvt-web
    Automated ArkTS code repair using retrieval-augmented generation, LLM-based test oracle synthesis, and structured benchmark evaluation for HarmonyOS development. Use when: 'fix this ArkTS error', 'repair HarmonyOS code', 'convert TypeScript to ArkTS', 'ArkTS compilation error', 'debug HarmonyOS component', 'generate tests for ArkTS code'.
    14 repo stars
  13. ▌
    Autonomous Business System Neuro Symbolic · ndpvt-web
    Design and implement neuro-symbolic business automation systems that combine LLM agents with predicate-logic programming and knowledge graphs. Applies the AUTOBUS architecture: tasks modeled as networks with pre/post conditions, enterprise data as logic facts, and AI agents that generate executable logic programs from natural-language instructions. Trigger phrases: - "Build a business process automation system" - "Create a neuro-symbolic workflow engine" - "Design a task orchestration system with logic constraints" - "Model business processes as knowledge graphs with AI agents" - "Implement an enterprise system that combines LLMs with rule engines" - "Build a system where AI agents generate executable business logic"
    14 repo stars
  14. ▌
    Autonomous Multi Agent AI High Throughput · ndpvt-web
    Build multi-agent AI systems for high-throughput scientific workflows with metacognitive self-assessment. Implements the Polymer Research Lifecycle (PRL) architecture: a Planner Agent decomposes complex scientific tasks into subtasks assigned to specialized domain agents (Research, Characterization, ML Model, Safety, Synthesis, Execution, Reporting), which produce consensus predictions with uncertainty estimates and continuously self-optimize via three-layer metacognitive reflection. Trigger phrases: - "Build a multi-agent pipeline for materials property prediction" - "Create a high-throughput screening system with agent consensus" - "Implement metacognitive self-assessment for an agent swarm" - "Design an autonomous scientific workflow with specialized agents" - "Set up a polymer informatics pipeline with uncertainty quantification" - "Orchestrate domain-specific agents for computational chemistry"
    14 repo stars
  15. ▌
    Bayesflow Probability Inference Framework · ndpvt-web
    Generate high-quality multi-step LLM workflows using Bayesian inference with parallel look-ahead rollouts and importance-weighted resampling. Use when: 'build a workflow for this task', 'generate an agent pipeline', 'create a multi-step LLM chain', 'optimize my prompt chain', 'design an agentic workflow', 'Bayesian workflow generation'.
    14 repo stars
  16. ▌
    C Mop Integrating Momentum Boundary Aware · ndpvt-web
    Optimize LLM system prompts iteratively using boundary-aware contrastive sampling and momentum-guided clustering from the C-MOP framework. Use when: 'optimize this prompt', 'improve my system prompt', 'evolve prompt for better accuracy', 'automatic prompt tuning', 'prompt optimization with examples', 'refine prompt using test cases'.
    14 repo stars
  17. ▌
    Cognitive Platform Engineering Autonomous · ndpvt-web
    Build autonomous cloud operations using a four-plane cognitive architecture (Sensing, Reasoning, Orchestration, Experience) with Kubernetes, Terraform, OPA, and ML anomaly detection. Use when: 'set up self-healing Kubernetes infrastructure', 'add anomaly detection to my cloud platform', 'create OPA policies for autonomous remediation', 'build a cognitive operations pipeline', 'implement intent-based infrastructure management', 'add intelligent auto-scaling with feedback loops'.
    14 repo stars
  18. ▌
    Commcp Multi Agent Coordination LLM Based · ndpvt-web
    Build decentralized multi-agent coordination systems using LLM-based communication calibrated with conformal prediction. Agents share only statistically reliable messages, reducing noise and redundancy. Use when: 'coordinate multiple agents with LLM messaging', 'build a multi-agent system with calibrated communication', 'reduce noisy messages between cooperating agents', 'implement conformal prediction for agent communication filtering', 'design decentralized agent coordination without a central controller', 'filter unreliable LLM-generated messages in multi-agent pipelines'.
    14 repo stars
  19. ▌
    Completing Missing Annotation Multi Agent · ndpvt-web
    Multi-agent debate framework for relevance assessment and annotation completion. Uses opposing-stance LLM agents with iterative critique to label query-document relevance, detect missing annotations, and escalate uncertain cases to humans. Triggers: 'assess relevance of documents to queries', 'find missing annotations in IR benchmark', 'debate-based relevance labeling', 'complete missing labels in dataset', 'multi-agent relevance assessment', 'evaluate document relevance with debate'
    14 repo stars
  20. ▌
    Discoverllm Executing Intents Discovering · ndpvt-web
    Help users discover and form their intents through adaptive diverge-converge interaction, rather than just asking clarifying questions. Use when the user gives a vague or open-ended request such as 'build me a website', 'write something creative', 'make a CLI tool', 'design a component', 'help me refactor this', or 'create a visualization'.
    14 repo stars
  21. ▌
    Diverge Diversity Enhanced RAG Open Ended · ndpvt-web
    Diversity-enhanced RAG for open-ended queries with multiple valid answers. Uses reflection-guided generation and memory-augmented iterative refinement to produce diverse, high-quality responses instead of collapsing to a single dominant answer. Triggers: 'give me diverse perspectives on', 'explore different viewpoints', 'brainstorm multiple approaches to', 'what are the different ways to', 'open-ended search with diversity', 'DIVERGE-style RAG search'
    14 repo stars
  22. ▌
    Empirical Mcts Continuous Agent Evolution · ndpvt-web
    Applies Empirical-MCTS dual-loop reasoning: structured tree search with persistent memory that accumulates experience across problems. Use when asked to 'solve this step by step with learning', 'use MCTS reasoning', 'try multiple approaches and remember what works', 'evolve your strategy as you go', 'search for the best solution systematically', or 'use tree search with memory'.
    14 repo stars
  23. ▌
    Error Taxonomy Guided Prompt Optimization · ndpvt-web
    Optimize LLM prompts by systematically collecting errors, building a taxonomy of failure modes, and augmenting prompts with targeted guidance for the most frequent error categories. Based on ETGPO (Singh et al., 2026). Trigger phrases: "optimize this prompt", "my prompt keeps failing", "improve prompt accuracy", "debug prompt errors", "fix LLM failures", "prompt isn't working well"
    14 repo stars
  24. ▌
    Graphseek Next Generation Graph Analytics · ndpvt-web
    Build LLM-powered graph analytics systems using the GraphSeek two-plane architecture: a Semantic Catalog for planning over graph schemas and operations, separated from deterministic database-grade query execution. Use when: 'query a property graph with natural language', 'build a graph analytics pipeline', 'connect LLM reasoning to Neo4j/Cypher', 'analyze a knowledge graph without writing queries', 'create a semantic layer for graph databases', 'multi-step graph traversal with LLM planning'.
    14 repo stars
  25. ▌
    Hallucination Resistant Security Planning · ndpvt-web
    Generate reliable incident response and security recovery plans using a generate-check-refine loop with consistency-based abstention. Reduces hallucinated security actions by validating LLM outputs against system constraints and lookahead predictions before committing. Trigger phrases: "plan incident response", "generate recovery plan", "security response plan", "analyze security logs", "incident recovery", "hallucination-resistant planning"
    14 repo stars
  26. ▌
    History Guided Iterative Visual Reasoning · ndpvt-web
    Apply the H-GIVR (History-Guided Iterative Visual Reasoning) framework for self-correcting multimodal reasoning. Uses iterative re-observation of images with accumulated answer history to dynamically correct errors and converge on accurate answers. Trigger phrases: - "analyze this image iteratively" - "use self-correction to answer this visual question" - "reason about this image step by step with verification" - "iteratively verify your answer about this image" - "use history-guided reasoning on this screenshot" - "double-check your visual analysis"
    14 repo stars
  27. ▌
    Hqp Sensitivity Aware Hybrid Quantization · ndpvt-web
    Apply the HQP framework to compress and accelerate PyTorch models for edge deployment using sensitivity-aware structural pruning followed by 8-bit post-training quantization. Trigger phrases: 'optimize model for edge', 'prune and quantize model', 'compress model for Jetson', 'reduce inference latency on edge device', 'hybrid quantization and pruning', 'deploy model to edge with size constraints'
    14 repo stars
  28. ▌
    Large Geolocation Extraction Humanitarian · ndpvt-web
    Extract and geocode location mentions from humanitarian and crisis texts using a two-step LLM pipeline: few-shot NER for toponym extraction followed by agent-based geocoding for coordinate resolution. Handles ambiguous place names, distinguishes literal from associative mentions, and reduces geographic bias. Use when: 'extract locations from crisis reports', 'geocode humanitarian text', 'find place names in disaster documents', 'NER for geographic entities in unstructured text', 'resolve ambiguous toponyms from news articles', 'geolocate mentions in humanitarian datasets'.
    14 repo stars
  29. ▌
    Legalmalr Multi Agent Query Understanding · ndpvt-web
    Multi-agent query reformulation and LLM reranking for retrieval over legal, regulatory, or domain-specific corpora. Use when building legal search systems, statute retrieval pipelines, or any RAG system where user queries are ambiguous, colloquial, or multi-issue. Triggers: 'build a legal search system', 'improve statute retrieval', 'multi-agent query reformulation', 'legal RAG pipeline', 'rerank legal search results', 'query understanding for retrieval'
    14 repo stars
  30. ▌
    Livibench Omnimodal Benchmark Interactive · ndpvt-web
    Build omnimodal benchmarks and evaluation pipelines for interactive video understanding (livestreams, real-time comments, multi-speaker audio). Applies LiViBench's multi-agent annotation workflow, seed-question-driven QA generation, and Video-to-Comment Retrieval (VCR) module design. Trigger phrases: 'build a livestream video benchmark', 'evaluate multimodal models on interactive video', 'create annotation pipeline for video QA', 'design a multi-agent video description system', 'implement comment retrieval for video understanding', 'build an omnimodal evaluation dataset'.
    14 repo stars
  31. ▌
    Realsec Bench Benchmark Evaluating Secure · ndpvt-web
    Evaluate and improve secure code generation using the RealSec-bench methodology: multi-stage vulnerability detection with CodeQL SAST scanning, inter-procedural data flow analysis, and composite security+correctness scoring via SecurePass@K. Applies lessons from real-world Java repository vulnerabilities spanning 19 CWE types. Use when: "audit this code for security vulnerabilities", "generate secure Java code", "check for CWE issues in this repo", "evaluate code security with CodeQL", "review this code for injection/crypto/data-flow vulnerabilities", "write secure code that avoids common weaknesses"
    14 repo stars
  32. ▌
    Rethinker Scientific Reasoning Rethinking · ndpvt-web
    Solve hard scientific and technical reasoning problems using the ReThinker Solver-Critic-Selector loop with confidence-gated rethinking. Use when the user says 'reason through this carefully', 'solve this hard problem', 'think step by step and verify', 'scientific reasoning', 'double-check your answer', or 'I need a confident answer to this technical question'.
    14 repo stars
  33. ▌
    Rethinking Scientific Modeling Physically · ndpvt-web
    Generate physics-consistent, simulation-executable structural engineering code using constraint-oriented alignment and verification-driven evaluation. Use when: 'generate an OpenSees model', 'create a structural analysis script', 'build a seismic-resistant frame model', 'write simulation-ready building code', 'check my structural model for physical consistency', 'debug my OpenSeesPy script'.
    14 repo stars
  34. ▌
    Semantic Aware Advanced Persistent Threat · ndpvt-web
    Build anomaly detection pipelines for Advanced Persistent Threat (APT) detection by encoding system logs into semantic embeddings with sentence-transformer LLMs and training autoencoders on reconstruction error. Trigger phrases: "detect APT in system logs", "semantic log anomaly detection", "autoencoder on log embeddings", "LLM-based threat detection", "provenance log analysis", "system log anomaly pipeline".
    14 repo stars
  35. ▌
    Sparseeval Evaluation Sparse Optimization · ndpvt-web
    Efficiently evaluate LLMs on benchmarks by selecting a small subset of anchor items via sparse optimization, reproducing full-benchmark rankings at a fraction of the cost. Use when: 'reduce evaluation cost for my LLM benchmark', 'select representative test items from a large dataset', 'rank models without running all benchmark samples', 'sparse subset selection for evaluation', 'find anchor items that represent my test suite', 'efficient model comparison on benchmarks'.
    14 repo stars
  36. ▌
    Swe Manager Selecting Synthesizing Golden · ndpvt-web
    Select and synthesize a golden proposal from multiple candidate fix strategies before writing code. Mirrors how technical managers deliberate on competing proposals by weighing scope, impact, risk, and each candidate's strengths and weaknesses. Use when: "compare these fix approaches", "which solution should we pick", "evaluate these proposals for the bug", "help me choose between these implementation strategies", "synthesize the best approach from these options", "review competing fixes before coding".
    14 repo stars
  37. ▌
    Symphony Synergistic Multi Agent Planning · ndpvt-web
    Orchestrate heterogeneous multi-agent MCTS planning for complex reasoning and search tasks. Uses a pool of diverse LLM-based agents with distinct roles (planner, executor, verifier, reflector) to generate diverse solution branches via Monte Carlo Tree Search, producing more robust plans than any single-agent approach. Trigger phrases: "use SYMPHONY planning", "multi-agent MCTS", "heterogeneous agent planning", "diverse rollout search", "synergistic planning agents", "multi-model tree search"
    14 repo stars
  38. ▌
    Toward Culturally Aligned Ontology Guided · ndpvt-web
    Ontology-guided multi-agent reasoning for culturally aligned LLM outputs. Use when building systems that must respect cultural values, when designing multi-agent pipelines with demographic grounding, or when implementing value-aware decision making. Triggers: 'culturally aligned generation', 'ontology-guided agents', 'value-persona reasoning', 'demographic-aware LLM', 'cultural sensitivity pipeline', 'WVS-style value alignment'
    14 repo stars
  39. ▌
    TS Debate Multimodal Collaborative Debate · ndpvt-web
    Zero-shot time series reasoning via modality-specialized multi-agent debate. Assigns dedicated text, visual, and numerical analyst agents to reason over temporal data, coordinated by structured debate and reviewer verification. Use when: 'analyze this time series data', 'classify this sensor signal', 'forecast this trend from context and charts', 'detect anomalies in this dataset', 'answer questions about this temporal data', 'reason over these multiple modalities together'.
    14 repo stars
  40. ▌
    Vihermes Graph Grounded Multihop Question · ndpvt-web
    Build graph-grounded multihop QA systems over regulatory and hierarchically structured documents. Combines vector similarity retrieval with knowledge graph traversal to answer questions requiring reasoning across multiple interconnected legal or policy documents. Triggers: 'build a regulatory QA system', 'multihop question answering over legal documents', 'graph-based retrieval for regulations', 'answer questions across multiple policy documents', 'healthcare regulation QA pipeline', 'cross-document reasoning with graph expansion'
    14 repo stars
  41. ▌
    Alertguardian Intelligent Alert Life Cycle · ndpvt-web
    Build intelligent alert lifecycle management systems for cloud infrastructure using graph-based denoising, RAG-powered summarization, and multi-agent rule refinement. Trigger phrases: - "reduce alert fatigue in our monitoring system" - "deduplicate and correlate alerts" - "summarize alerts for on-call engineers" - "refine our alerting rules automatically" - "build an alert denoising pipeline" - "too many alerts, help me triage"
    14 repo stars
  42. ▌
    Beyond Instrumental Substitutive Paradigms · ndpvt-web
    Audit and diagnose cultural bias artifacts in LLM-powered applications using the Machine Culture framework. Detects Cultural Reversal (language-culture misalignment), Service Persona Camouflage (RLHF-induced affective flattening), and superposition-driven inconsistencies. Use when: 'audit my prompts for cultural bias', 'why does my chatbot respond differently in Chinese vs English', 'detect persona camouflage in model outputs', 'cross-cultural prompt testing', 'diagnose RLHF flattening in my AI product', 'multilingual cultural consistency check'.
    14 repo stars
  43. ▌
    C2rope Causal Continuous Rotary Positional · ndpvt-web
    Implement C²RoPE (Causal Continuous Rotary Positional Encoding) for multimodal transformers that process 2D/3D visual data alongside text. Replaces standard 1D RoPE with a triplet (m, x, y) positional index and Chebyshev causal masking to preserve spatial locality in vision-language models. Trigger phrases: - "implement C2RoPE positional encoding" - "fix spatial locality loss in vision-language RoPE" - "add 2D-aware rotary embeddings for image tokens" - "implement Chebyshev causal masking for visual attention" - "modify RoPE for multimodal 3D reasoning" - "spatially-aware positional encoding for multi-view images"
    14 repo stars
  44. ▌
    Co Redteam Orchestrated Security Discovery · ndpvt-web
    Multi-agent security vulnerability discovery and exploitation using Co-RedTeam's orchestrated workflow. Decomposes security analysis into coordinated discovery and exploitation stages with execution-grounded iterative reasoning and layered memory. Use when: 'find vulnerabilities in this codebase', 'red team this application', 'security audit this project', 'exploit this vulnerability', 'penetration test this service', 'analyze this code for security flaws'.
    14 repo stars
  45. ▌
    Craft Calibrated Reasoning Answer Faithful · ndpvt-web
    Apply CRAFT (Calibrated Reasoning with Answer-Faithful Traces) for multi-hop question answering with verified reasoning chains. Use when: 'answer a complex question from multiple documents', 'build a faithful multi-hop QA pipeline', 'reason over retrieved passages with citations', 'verify reasoning chain faithfulness', 'structured RAG with auditable traces', 'multi-step question decomposition with evidence grounding'
    14 repo stars
  46. ▌
    Eliciting Least To Most Reasoning Phishing · ndpvt-web
    Detect phishing URLs using Least-to-Most iterative decomposition with answer sensitivity scoring. Triggers: 'analyze this URL for phishing', 'is this URL safe', 'check URL for phishing indicators', 'phishing detection', 'classify this URL as phishing or benign', 'scan URL for suspicious patterns'
    14 repo stars
  47. ▌
    Failure Aware Enhancements Code Generation · ndpvt-web
    Diagnose why generated code fails and apply the right fix strategy (self-critique, RAG, multi-model, or progressive prompting) based on a data-driven decision framework from empirical research on 25 GitHub projects. Trigger phrases: "my generated code doesn't work", "fix this code generation failure", "why does this code keep failing", "help me debug LLM-generated code", "improve code generation quality", "the AI-generated code is wrong"
    14 repo stars
  48. ▌
    From Code Centric Concept Centric Teaching · ndpvt-web
    Generate LLM-assisted coding labs that teach concepts through 'Vibe Coding' — producing working code paired with mandatory conceptual reflection, prompt logging, and critical thinking assessments. Use when: 'create a vibe coding lab for transformers', 'design an NLP exercise with reflection questions', 'build a concept-focused coding tutorial', 'generate a lab that teaches X concept not just syntax', 'create a coding assignment with critical reflection', 'design a learn-by-prompting exercise'.
    14 repo stars
  49. ▌
    From Prompt Response Goal Directed Systems · ndpvt-web
    Design production-grade agentic AI architectures with separated cognition/execution layers, typed tool interfaces, multi-agent topologies, and enterprise hardening. Use when: 'design an agent system', 'build a multi-agent architecture', 'add governance to my AI pipeline', 'harden my LLM agent for production', 'create a tool registry for agents', 'architect agent-to-agent coordination'.
    14 repo stars
  50. ▌
    Gisa Benchmark General Information Seeking · ndpvt-web
    Build structured information-seeking agents that decompose complex queries into multi-turn search-and-browse workflows, aggregate results from multiple web sources, and return answers in typed structured formats (items, sets, lists, tables). Applies the GISA benchmark's ReAct-based agent architecture and evaluation methodology. Trigger phrases: "build an information-seeking agent", "search agent pipeline", "multi-turn web research agent", "structured web search workflow", "aggregate information from multiple sources", "web research with structured output"
    14 repo stars
  51. ▌
    Graphagents Knowledge Graph Guided Agentic · ndpvt-web
    Build multi-agent pipelines that use knowledge graphs to guide LLM reasoning across domains. Agents specialize in problem decomposition, evidence retrieval, parameter extraction, graph traversal, and hypothesis synthesis. Use when: 'build a knowledge graph agent pipeline', 'find cross-domain material substitutes', 'multi-agent graph traversal system', 'knowledge graph guided search', 'graph-based RAG with specialized agents', 'cross-domain design exploration with KG'.
    14 repo stars
  52. ▌
    Kg Craft Knowledge Graph Based Contrastive · ndpvt-web
    Fact-check claims using knowledge graph-based contrastive reasoning. Constructs a KG from claims and evidence sources, generates contrastive questions ("Why X rather than Y?") grounded in the KG structure, distills evidence through targeted Q&A, and synthesizes a summary for veracity assessment. Trigger phrases: "fact-check this claim", "verify this statement against sources", "build a knowledge graph for claim verification", "contrastive reasoning for fact-checking", "check if this claim is true using these reports", "KG-CRAFT pipeline"
    14 repo stars
  53. ▌
    Llms As Orchestrators Constraint Compliant · ndpvt-web
    Build constraint-compliant multi-objective recommendation systems using a dual-agent architecture coordinated by an LLM. Implements the DualAgent-Rec framework: an Exploitation Agent that optimizes accuracy under hard business constraints and an Exploration Agent that searches unconstrained Pareto space for diversity, with adaptive epsilon-relaxation guaranteeing 100% constraint satisfaction. Trigger phrases: - "Build a recommendation system with hard constraints" - "Multi-objective optimization with constraint satisfaction" - "DualAgent-Rec" or "dual-agent recommendation" - "Constrained Pareto optimization for recommendations" - "Fairness and coverage constraints in recommendations" - "LLM-coordinated multi-agent optimization"
    14 repo stars
  54. ▌
    Multivis Agent Multi Agent Framework Logic · ndpvt-web
    Build reliable multi-agent data visualization pipelines with logic rule constraints. Use when: 'generate a chart from this database', 'create a visualization matching this reference image', 'refine this chart based on feedback', 'build a multi-agent visualization system', 'convert this matplotlib code to altair', 'visualize SQL query results with style references'.
    14 repo stars
  55. ▌
    Qrs Rule Synthesizing Neuro Symbolic Triad · ndpvt-web
    Autonomous vulnerability discovery using the QRS (Query, Review, Sanitize) neuro-symbolic triad. Generates CodeQL queries from CWE schemas, validates findings through semantic reasoning, and confirms exploitability via PoC synthesis. Use when: 'scan this Python package for vulnerabilities', 'generate CodeQL queries for CWE-078', 'review these SAST findings for false positives', 'find injection vulnerabilities in this codebase', 'write a taint-tracking query for path traversal', 'triage these security alerts'.
    14 repo stars
  56. ▌
    Solagent Specialized Multi Agent Framework · ndpvt-web
    Generate secure, functionally correct Solidity smart contracts using a dual-loop refinement process: an inner loop that compiles and tests with Forge until all tests pass, and an outer loop that runs Slither static analysis to eliminate security vulnerabilities. Triggers: 'write a Solidity contract', 'generate smart contract', 'secure Solidity code', 'fix smart contract vulnerabilities', 'audit and fix Solidity', 'create ERC-20/ERC-721 contract with tests'
    14 repo stars
  57. ▌
    Structured Context Engineering File Native · ndpvt-web
    Structure database schemas and structured data as file-native context for LLM agent operations. Applies evidence-based format selection, domain-partitioned file layouts, and architecture-aware context engineering to maximize accuracy when agents generate SQL, query APIs, or navigate large structured systems. Use this skill when: - "Generate SQL against a database with many tables" - "Set up schema context for an agent pipeline" - "Organize database metadata files for LLM consumption" - "Help me structure context for a text-to-SQL system" - "Scale my agent to handle thousands of tables" - "Choose the best format for feeding schema info to an LLM"
    14 repo stars
  58. ▌
    Synthagent Multi Agent Framework Realistic · ndpvt-web
    Build multi-agent pipelines that generate realistic synthetic patient profiles by integrating epidemiological data, medical claims, literature evidence, and personality models. Use when asked to 'simulate patients', 'generate synthetic medical data', 'build a patient simulation pipeline', 'create virtual patient cohorts', 'model disease progression with agents', or 'synthesize clinical profiles'.
    14 repo stars
  59. ▌
    Think Augmented Function Calling Improving · ndpvt-web
    Improve LLM function/tool calling accuracy by injecting explicit "think" reasoning parameters into function schemas before the model generates arguments. Based on the TAFC framework (arXiv:2601.18282). Use this skill when: - "Add reasoning to my tool definitions" - "My agent is passing wrong parameters to functions" - "Improve function calling accuracy for complex APIs" - "Debug why my LLM agent picks incorrect argument values" - "Make my tool calls more reliable with chain-of-thought" - "Augment my OpenAI function schemas with reasoning"
    14 repo stars
  60. ▌
    Thinktank Me Multi Expert Framework Middle · ndpvt-web
    Build multi-expert forecasting systems where specialized LLM agents collaborate through routing and aggregation to predict complex events. Use when asked to: 'build a multi-expert prediction system', 'create specialized agents that collaborate on forecasting', 'implement expert routing for predictions', 'aggregate predictions from multiple specialized models', 'design a think-tank style multi-agent architecture', 'forecast geopolitical or temporal events with multiple experts'.
    14 repo stars
  61. ▌
    Toolweaver Weaving Collaborative Semantics · ndpvt-web
    Design scalable tool retrieval systems using hierarchical code tokenization that captures collaborative tool semantics. Use when: 'build a tool registry with hierarchical codes', 'scale tool selection for thousands of APIs', 'encode tool co-usage patterns', 'design a generative tool retrieval pipeline', 'organize APIs into a searchable hierarchy', 'implement constrained decoding for tool selection'.
    14 repo stars
  62. ▌
    Unit Based Agent Semi Cascaded Full Duplex · ndpvt-web
    Build full-duplex voice dialogue systems using unit-based agent decomposition and semi-cascaded pipelines. Trigger phrases: 'build a full-duplex dialogue system', 'implement voice interaction with interruption handling', 'create a real-time conversational agent', 'design a semi-cascaded speech pipeline', 'add backchannel and turn-taking to a voice agent', 'unit-based dialogue processing'
    14 repo stars
  63. ▌
    Xai Clip Roi Guided Perturbation Framework · ndpvt-web
    Build ROI-guided perturbation pipelines for explainable medical image segmentation using CLIP embeddings. Generates boundary-aware saliency maps by restricting perturbations to anatomically meaningful regions identified via vision-language models. Use when: 'explain medical segmentation predictions', 'build XAI pipeline for CT/MRI', 'generate saliency maps for organ segmentation', 'reduce perturbation cost for explainability', 'CLIP-guided occlusion sensitivity', 'ROI-aware LIME/RISE for medical images'.
    14 repo stars
  64. ▌
    Agentstepper Interactive Debugging Software · ndpvt-web
    Interactive debugging of LLM-powered software development agents using structured trajectory analysis, stepwise execution, and live editing of prompts/tool calls. Use when: 'debug my agent', 'why did the agent do that', 'trace agent execution', 'step through agent actions', 'inspect agent trajectory', 'agent is producing wrong output'.
    14 repo stars
  65. ▌
    Deepread Document Structure Aware Reasoning · ndpvt-web
    Structure-aware document reasoning that converts PDFs/long documents into hierarchically indexed paragraphs with coordinate metadata, then uses a dual-tool "locate then read" strategy (Retrieve + ReadSection) to answer complex questions requiring evidence scattered across distant document sections. Trigger phrases: - "analyze this PDF and answer questions about it" - "find evidence across sections of this document" - "search this long document for specific information" - "extract structured answers from this report" - "answer questions about this paper/filing/manual" - "build a document QA pipeline"
    14 repo stars
  66. ▌
    Dial Summer Structured Evaluation Framework · ndpvt-web
    Evaluate dialogue summaries using the DIAL-SUMMER hierarchical error taxonomy. Detects 10 fine-grained error types across two levels: dialogue-level (speaker/turn structure) and within-turn-level (semantic content). Use when: 'evaluate this meeting summary', 'check this dialogue summary for errors', 'find hallucinations in this conversation summary', 'audit this chat transcript summary', 'grade this call center summary', 'detect speaker misattribution in this summary'.
    14 repo stars
  67. ▌
    Evoconfig Self Evolving Multi Agent Systems · ndpvt-web
    Autonomous environment configuration using multi-agent diagnosis and self-evolving error repair. Use when: 'set up the dev environment for this repo', 'configure dependencies and get tests passing', 'debug why my environment build fails', 'create a Dockerfile for this project', 'fix my broken Python environment', 'auto-configure this repository so tests run'.
    14 repo stars
  68. ▌
    Fullstack Agent Enhancing Agentic Fullstack · ndpvt-web
    Build production-grade full-stack web applications using a three-agent pipeline (Planning, Backend, Frontend) with development-oriented testing and structured debugging. Triggers: 'build a full-stack app', 'create a web app with backend and database', 'full-stack website with API and database', 'build me a CRUD app', 'create a web application with user authentication and data storage', 'scaffold a complete web project with frontend and backend'
    14 repo stars
  69. ▌
    Gutenocr Grounded Vision Language Front End · ndpvt-web
    Build grounded OCR pipelines using GutenOCR's prompt-based interface for reading, detection, and spatial grounding on documents. Use when: 'extract text with bounding boxes from a PDF', 'find where a phrase appears in a scanned document', 'build a document OCR pipeline with spatial coordinates', 'detect text regions in business documents', 'locate specific content in a scientific paper image', 'read text from a cropped region of a document'.
    14 repo stars
  70. ▌
    Infa Guard Mitigating Malicious Propagation · ndpvt-web
    Implement infection-aware security for LLM multi-agent systems using INFA-Guard's three-category detection (benign/attacker/infected), topological constraint analysis, and graduated remediation (replace attackers, rehabilitate infected agents). Use when: "secure my multi-agent system", "detect prompt injection in agent swarm", "prevent malicious propagation between agents", "add infection-aware guard to MAS", "harden agent communication topology", "defend against agent-to-agent attacks".
    14 repo stars
  71. ▌
    Locomo Plus Beyond Factual Cognitive Memory · ndpvt-web
    Build and evaluate cognitive memory systems for LLM dialogue agents that retain implicit user constraints (state, goals, values, causal context) across long conversations -- not just explicit facts. Use when: 'design a memory system for my chatbot', 'evaluate agent long-term memory', 'handle implicit user preferences in conversations', 'build constraint-aware dialogue', 'test cognitive recall in my agent', 'improve memory beyond factual retrieval'.
    14 repo stars
  72. ▌
    Mermaid Memory Enhanced Retrieval Reasoning · ndpvt-web
    Memory-enhanced multi-agent retrieval and reasoning for veracity assessment and fact-checking. Use when: 'verify this claim', 'fact-check these statements', 'check if this is true', 'assess the veracity of', 'cross-reference these claims', 'build a fact-checking pipeline'.
    14 repo stars
  73. ▌
    Mulvul Retrieval Augmented Multi Agent Code · ndpvt-web
    Multi-agent vulnerability detection using coarse-to-fine routing, contrastive retrieval, and cross-model prompt evolution. Use when: 'scan this code for vulnerabilities', 'detect CWE weaknesses in this function', 'security audit this C/C++ code', 'find buffer overflows and injection flaws', 'analyze this code for common weakness enumerations', 'run a multi-agent vulnerability check on this codebase'.
    14 repo stars
  74. ▌
    Profinfer Ebpf Based Fine Grained Inference · ndpvt-web
    Profile and diagnose LLM inference engines (llama.cpp and similar GGML-based runtimes) using eBPF uprobes for non-intrusive, operator-level performance analysis. Trigger phrases: 'profile llama.cpp inference', 'eBPF LLM profiling', 'diagnose inference bottleneck', 'trace GGML operators', 'is my inference memory-bound or compute-bound', 'profile MoE routing overhead'
    14 repo stars
  75. ▌
    Rethinking Reranker Boundary Aware Evidence · ndpvt-web
    Implement boundary-aware evidence selection for RAG systems using the BAR-RAG technique. Replaces relevance-only reranking with difficulty-calibrated evidence selection that targets the generator's competence boundary -- passages that are challenging yet solvable. Use when: "build a robust RAG pipeline", "improve reranking for noisy retrieval", "select better evidence for question answering", "make my RAG system handle noisy documents", "implement boundary-aware reranking", "calibrate evidence difficulty for my LLM generator".
    14 repo stars
  76. ▌
    Scratcheval Multimodal Evaluation Framework · ndpvt-web
    Evaluate, debug, and repair block-based Scratch programs using a three-layer executable protocol (VM execution, block-level edit distance, explanation rubrics). Use when: 'debug this Scratch project', 'analyze Scratch program logic', 'find bugs in my Scratch code', 'evaluate Scratch repair quality', 'convert Scratch blocks to analyzable AST', 'fix concurrency issues in Scratch sprites'.
    14 repo stars
  77. ▌
    Self Hinting Enhance Reinforcement Learning · ndpvt-web
    Apply the SAGE self-hinting technique to improve LLM problem-solving by generating graduated hints that boost solution diversity and prevent reasoning collapse. Use when: 'help me solve this hard math problem step by step', 'generate hints for this coding challenge', 'break down this complex problem with progressive clues', 'design a self-hinting prompt pipeline', 'create a hint curriculum for training data', 'scaffold this reasoning task with decomposition hints'.
    14 repo stars
  78. ▌
    SQL Trail Multi Turn Reinforcement Learning · ndpvt-web
    Iterative multi-turn Text-to-SQL generation using reason-execute-observe loops with execution feedback. Instead of writing SQL in one shot, explores the schema, executes intermediate queries, reads error messages and result previews, and refines until correct. Use when: 'write a SQL query for this database', 'help me query this schema', 'debug this SQL query', 'convert this question to SQL', 'analyze this database and answer', 'fix my failing SQL'.
    14 repo stars
  79. ▌
    Standardizing Longitudinal Radiology Report · ndpvt-web
    Build LLM-based pipelines that automatically detect and classify longitudinal (temporal) changes in radiology reports. Use when the user mentions 'radiology report annotation', 'longitudinal report evaluation', 'temporal change detection in medical text', 'disease progression extraction', 'radiology NLP pipeline', or 'benchmark radiology report generation'.
    14 repo stars
  80. ▌
    Steereval Framework Evaluating Steerability · ndpvt-web
    Evaluate and improve the steerability of natural-language-profile-based recommender systems using the SteerEval framework. Build evaluation pipelines that measure whether recommendation engines actually respond to user steering commands (preference edits). Use when: 'evaluate my recommender's steerability', 'test if profile edits change recommendations', 'build a steerable recommendation system', 'measure how well my rec system follows user preferences', 'benchmark natural language profile steering', 'audit recommendation controllability'.
    14 repo stars
  81. ▌
    Sycoeval Em Sycophancy Evaluation Simulated · ndpvt-web
    Build multi-agent adversarial simulations to evaluate LLM sycophancy and policy compliance under social pressure. Use when asked to 'test LLM robustness against manipulation', 'evaluate sycophancy in AI systems', 'build adversarial red-team simulations', 'stress-test LLM guardrails with persuasion', 'create multi-agent safety benchmarks', or 'simulate social pressure attacks on AI assistants'.
    14 repo stars
  82. ▌
    The Compliance Paradox Semantic Instruction · ndpvt-web
    Detect and defend against adversarial prompt injections hidden in code submissions that exploit LLM instruction-following to manipulate automated evaluation. Applies the SPACI/AST-ASIP framework to audit code for hidden directives in comments, identifiers, and dead code. Use when: 'audit this code for hidden prompt injections', 'is this submission trying to manipulate the grader', 'harden my LLM code evaluator', 'check for adversarial directives in student code', 'detect compliance paradox attacks', 'scan AST trivia nodes for injections'.
    14 repo stars
  83. ▌
    Triplay Rl Tri Role Self Play Reinforcement · ndpvt-web
    Apply the TriPlay-RL tri-role adversarial self-play framework to systematically red-team, harden, and evaluate LLM-powered applications for safety. Trigger phrases: 'red-team my LLM app', 'adversarial safety testing', 'tri-role safety audit', 'harden my chatbot against jailbreaks', 'evaluate LLM safety alignment', 'self-play safety loop'
    14 repo stars
  84. ▌
    Vtc R1 Vision Text Compression Long Context · ndpvt-web
    Implement VTC-R1 vision-text compression for efficient long-context reasoning. Renders intermediate reasoning segments into images as 'optical memory' fed back into vision-language models, achieving 3.4x token compression and 2.7x latency speedup. Use when: 'compress long reasoning chains', 'implement optical memory for VLMs', 'build iterative vision-text reasoning loop', 'reduce inference tokens with image rendering', 'VTC-R1 reasoning pipeline', 'text-to-image compression for LLM reasoning'.
    14 repo stars
  85. ▌
    Agent Fence Mapping Security Vulnerabilities · ndpvt-web
    Audit LLM agent systems for trust-boundary security vulnerabilities using the AgentFence taxonomy of 14 attack classes across planning, memory, retrieval, tool use, and delegation. Produces trace-auditable security reports with mean security break rates (MSBR) per attack surface. Use when: "audit my agent for security vulnerabilities", "check agent trust boundaries", "find delegation attacks in my agent code", "map security risks in my LangGraph/CrewAI pipeline", "evaluate agent architecture security", "run AgentFence analysis on this agent system".
    14 repo stars
  86. ▌
    Agenticsimlaw Juvenile Courtroom Multi Agent · ndpvt-web
    Structured multi-agent courtroom debate for explainable high-stakes tabular decisions. Use when: 'set up a multi-agent debate for this prediction', 'use adversarial agents to classify this table row', 'build a courtroom-style reasoning pipeline', 'create prosecutor/defense/judge agents for this decision', 'explain this tabular prediction with a debate', 'run a structured adversarial analysis on this dataset'.
    14 repo stars
  87. ▌
    Analyticsgpt Workflow Scientometric Question · ndpvt-web
    Build sequential LLM pipelines for scientometric question answering over academic databases. Decomposes meta-scientific queries into entity recognition, multi-step planning, parallel data retrieval, and analytical synthesis. Use when: 'build a scientometric QA system', 'answer questions about research impact', 'query academic publication databases with natural language', 'analyze citation metrics for institutions or authors', 'create a pipeline for science-of-science questions', 'implement RAG over scholarly metadata'.
    14 repo stars
  88. ▌
    Attn Gs Attention Guided Context Compression · ndpvt-web
    Compress long user contexts (profiles, histories, documents) into concise, high-quality summaries using attention-guided importance marking. Use when asked to: 'compress this user profile', 'shorten this context for the prompt', 'reduce token usage for personalization', 'summarize interaction history preserving key signals', 'fit this long context into a token budget', 'extract the most relevant parts of this document for a query'.
    14 repo stars
  89. ▌
    Beyond Function Level Analysis Context Aware · ndpvt-web
    Inter-procedural vulnerability detection using context-aware reasoning. Analyzes functions alongside their callers, callees, and global state to find vulnerabilities that single-function analysis misses. Uses code property graph traversal, security-focused context profiling, relevance scoring, and structured reasoning traces. Trigger phrases: - "Check this code for vulnerabilities across function boundaries" - "Analyze this function with its callers and callees for security issues" - "Find inter-procedural vulnerabilities in this codebase" - "Review this code for vulnerabilities that depend on how it's called" - "Do a deep security audit with cross-function context" - "Analyze whether this function is safe given how callers use it"
    14 repo stars
  90. ▌
    Cognitively Diverse Multiple Choice Question · ndpvt-web
    Generate high-quality multiple-choice questions at controlled cognitive levels using the ReQUESTA multi-agent framework. Decomposes MCQ authoring into planning, generation, evaluation, and post-processing stages with specialized agents targeting text-based (recall), inferential (synthesis), and main idea (abstraction) comprehension. Trigger phrases: "generate MCQs from this text", "create quiz questions at different difficulty levels", "make multiple choice questions for this reading", "build an assessment from this passage", "create comprehension questions", "generate exam items from this content"
    14 repo stars
  91. ▌
    Dart Diffusion Inspired Speculative Decoding · ndpvt-web
    Set up and use DART (Diffusion-Inspired Speculative Decoding) for fast LLM inference. DART replaces autoregressive draft models with parallel masked-position prediction using a single transformer layer, combined with N-gram-enforced tree pruning. Triggers: 'speed up LLM inference with DART', 'set up speculative decoding with DART', 'integrate DART for faster generation', 'configure DART draft model', 'compare DART vs EAGLE3 speculative decoding', 'optimize LLM serving latency with parallel drafting'.
    14 repo stars
  92. ▌
    Davinci Agency Unlocking Long Horizon Agency · ndpvt-web
    Decompose complex, long-horizon coding tasks into PR-like chains of verifiable subtasks with cross-stage dependency tracking and iterative refinement. Use when: 'break this feature into PRs', 'plan a multi-step implementation', 'decompose this project into verifiable stages', 'help me build this feature incrementally with tests at each step', 'create a chain of PRs for this task', 'plan an implementation with bug-fix iterations'.
    14 repo stars
  93. ▌
    Evaluating Enhancing Vulnerability Reasoning · ndpvt-web
    Perform DAG-structured vulnerability reasoning on code, modeling causal dependencies between code facts instead of linear chain-of-thought. Use when asked to: 'analyze this code for vulnerabilities', 'explain why this code is vulnerable', 'trace the root cause of this security bug', 'review this function for memory safety issues', 'is this code exploitable and why', 'reason about the security of this code path'.
    14 repo stars
  94. ▌
    Fs Researcher Test Time Scaling Long Horizon · ndpvt-web
    File-system-based dual-agent deep research framework that scales beyond context windows. Separates evidence gathering (Context Builder) from report writing (Report Writer) using a persistent hierarchical knowledge base on disk. Use this skill when the user says: "research this topic in depth", "write a comprehensive report on X", "do deep research about Y", "investigate and write up Z thoroughly", "build a knowledge base and then write a report", "scale up research quality with more rounds".
    14 repo stars
  95. ▌
    Gamedevbench Evaluating Agentic Capabilities · ndpvt-web
    Agentic game development with visual feedback loops for Godot Engine projects. Applies the GameDevBench methodology: navigating scene hierarchies, editing multimodal assets (sprites, shaders, animations), and using screenshot/video feedback to verify changes visually. Trigger phrases: 'build a game in Godot', 'fix my Godot scene', 'add animation to my game character', 'edit this shader effect', 'set up sprite animations from a spritesheet', 'create a game UI with Godot'
    14 repo stars
  96. ▌
    Longcat Flash Thinking 2601 Technical Report · ndpvt-web
    Build robust multi-tool agentic pipelines with noise-aware execution, parallel reasoning, and environment scaling patterns from the LongCat-Flash-Thinking architecture. Use when: 'build a multi-tool agent pipeline', 'make my agent robust to API failures', 'scale tool orchestration across domains', 'add parallel reasoning to my agent', 'handle noisy tool responses', 'design a fault-tolerant agentic workflow'.
    14 repo stars
  97. ▌
    Mas Prove Understanding Process Verification · ndpvt-web
    Design and implement process verification for multi-agent LLM systems. Add intermediate-step evaluation to multi-agent workflows using LLM-as-a-Judge, reward models, or process reward models at agent-level or iteration-level granularity. Trigger phrases: 'add verification to my multi-agent system', 'evaluate intermediate steps in my agent pipeline', 'process verification for MAS', 'verify agent reasoning steps', 'judge agent outputs in a multi-agent workflow', 'add quality checks between agent steps'.
    14 repo stars
  98. ▌
    Mitigating Conversational Inertia Multi Turn · ndpvt-web
    Detect and break conversational inertia in multi-turn agent interactions — where an LLM repeats its own prior actions as implicit few-shot examples instead of exploring alternatives. Apply Context Preference Learning and context management to improve agentic task performance. Use when: 'my agent keeps repeating the same action', 'break out of action loops', 'agent is stuck in a loop', 'improve multi-turn agent exploration', 'balance exploration and exploitation in agent', 'reduce imitation bias in tool-calling agent'.
    14 repo stars
  99. ▌
    Mmr Bench Comprehensive Benchmark Multimodal · ndpvt-web
    Build cost-aware multimodal LLM routing systems that select the best model per query based on input signals, budget constraints, and task type. Use when the user says 'route queries to different models', 'select the cheapest model that works', 'build a model router', 'multimodal model selection', 'cost-optimize LLM inference', or 'Pareto-optimal model dispatch'.
    14 repo stars
  100. ▌
    Multi Agent Constraint Factorization Reveals · ndpvt-web
    Orchestrate multi-agent LLM pipelines using constraint factorization -- decomposing complex requirements into separate constraint-enforcement agents that iteratively project a shared solution toward feasibility. Use when: 'build a multi-agent pipeline', 'decompose this problem into agent constraints', 'set up constraint-based agent orchestration', 'factorize requirements across agents', 'iterative multi-agent refinement loop', 'cyclic projection agent workflow'.
    14 repo stars