Classify textual tasks into complexity bands and propose abstract model labels.
npx skills add https://github.com/JotJunior/cstk --skill model-selector-jotjunior
Please help me install this Agent Skill. Skill: model-selector Source: https://github.com/JotJunior/cstk/tree/main/global/skills/model-selector Command: npx skills add https://github.com/JotJunior/cstk --skill model-selector-jotjunior
Classifies a textual task into complexity bands (rasa, media, profunda) and proposes an abstract model label (haiku, sonnet, opus, or manter-atual) to guide model selection for the next step in a pipeline.
Feed a textual task description to the classifier to obtain a model suggestion and score.
High-intent search queries and answers about installing and using this skill.
Task complexity classification assigns textual inputs to bands like rasa, media, or profunda, proposing a model label such as haiku, sonnet, or opus to guide selection. It uses a deterministic signal catalog to score the input and output a structured Markdown justification.
Automated model selection uses a deterministic classifier to evaluate task descriptions and suggest a cost-effective model tier. It outputs a Markdown block with the suggested model, score, and justification, designed for integration by an orchestrator before calling heavier sub-agents.
Yes, deterministic task classification runs offline relying exclusively on POSIX tools like awk, grep, and sed. It reads a local signal catalog and processes input strings up to 4096 characters without requiring network access or external dependencies.
The model selector is designed for orchestrator integration, providing a standardized Markdown output with SuggestedModel, Score, Justification, and Alternative sections. This format enables automated parsing to route tasks to appropriate sub-agents based on complexity bands.
Signal catalog classification is limited to input strings of 4096 characters and processes a single task at a time. It relies on a static local references file, meaning the signal catalog must be manually maintained to reflect new task patterns or complexity indicators accurately.
Deterministic signal-based classification uses fixed pattern matching with POSIX tools to assign complexity tiers, while dynamic routing typically consumes tokens for inference. The deterministic approach offers reproducible scores and offline operation but lacks adaptive contextual understanding of novel inputs.
nopperabboValidate Jenkins pipeline and shared library syntax, structure, and security.
DarkArty07Decompose complex multi-agent goals into kanban task graphs with dependencies and specialist assignees.
getspurStandardize blocker, scope drift, and architectural risk reporting in multi-agent coding environments.
EndUser123Discover LLM models dynamically and route requests by measured performance across providers.
