Recommend lightweight skills or heavyweight agents for given tasks.
npx skills add https://github.com/0xDarkMatter/claude-mods --skill tool-discovery
Please help me install this Agent Skill. Skill: tool-discovery Source: https://github.com/0xDarkMatter/claude-mods/tree/main/skills/tool-discovery Command: npx skills add https://github.com/0xDarkMatter/claude-mods --skill tool-discovery
This Skill helps you quickly identify whether to use a lightweight SKILL tool or a heavyweight AGENT for a given task, reducing decision fatigue and speeding up automation pipelines.
Ask Claude: "For analyzing this dataset, which tool or agent should I use and why?"
High-intent search queries and answers about installing and using this skill.
Use a skill for lightweight, single-purpose tasks like lookups or pattern matching; use an agent for complex reasoning, multi-step decisions, or tasks requiring orchestration across multiple tools. Skills are fast and focused; agents handle nuanced decision flows.
Skills are optimized for specific, narrow tasks with minimal overhead—ideal for file searches or code analysis. Agents are heavier-weight systems designed for complex reasoning, task orchestration, and decisions requiring context from multiple sources.
Choose an agent when your task requires reasoning across multiple inputs, conditional logic, or orchestrating several operations together. Skills excel at isolated, deterministic tasks; agents handle ambiguity and multi-step workflows.
Yes. Skills handle API design guidance, data processing patterns, and decision logic through reference catalogs and triggered recommendations. They work across task types including lookups, pattern matching, code analysis, and design decisions.
Tool recommendations are capped at 2–3 actionable pairings per task. This focused approach reduces decision fatigue and ensures suggestions are specific and immediately usable rather than overwhelming.
It provides decision logic, reference catalogs for common task patterns, and clear triggers for when to inject a skill versus an agent. Guidance covers lookups, reasoning tasks, code analysis, and API design scenarios.
NachoFLizaurTriage post-feature requests into lightweight, partial restart, or full restart paths.
farmountainCalibrate AI agent confidence scores against actual outcomes.
opensage-agentOrchestrate multi-agent systems with ensemble, pool, and retry patterns.
PostalocityAutomates system design decisions for complex software architectures.
