Generate and evaluate 2–4 candidate approaches for sub-tasks.
npx skills add https://github.com/dgroch/metacognition --skill resource-selection
Please help me install this Agent Skill. Skill: resource-selection Source: https://github.com/dgroch/metacognition/tree/main/skills/resource-selection Command: npx skills add https://github.com/dgroch/metacognition --skill resource-selection
Deliberate tool selection to avoid defaulting to a single solution when multiple viable approaches exist, reducing waste and errors.
Generate 2–4 candidate approaches, evaluate them against key criteria, and select the best tool to complete the task.
High-intent search queries and answers about installing and using this skill.
To pick the optimal tool, generate 2–4 candidate approaches by consulting capability inventories and environment constraints, then evaluate them against fitness, efficiency, reliability, and environment fit to select the best option.
Deliberate tool selection is an AI agent decision-making process that proactively generates multiple viable approaches and evaluates them against defined criteria to avoid defaulting to a single solution, reducing waste and errors.
Evaluate candidate approaches by comparing options across fitness, efficiency, reliability, environment fit, and reversibility, leveraging capability inventories and failure history to guide decisions and safeguard against poor choices.
You need deliberate resource selection when multiple viable approaches exist for a sub-task and you must avoid defaulting to a single solution, ensuring the chosen tool aligns with environment constraints and failure history.
Resource selection coordinates with failure recovery, environment model, and task decomposition skills to maintain coherence, providing a commitment protocol with recovery paths if a step fails.
Defaulting to a single tool for sub-tasks limits your ability to adapt to environment constraints, ignores failure history, and increases waste and errors by bypassing an evaluation framework that checks fitness, efficiency, and reliability.
fabiomilennials1234-a11yTranslate Conductor briefs into parallel sub-task prompts with acceptance criteria.
hiddink-aiDetect repetitive failure loops in agent task execution via tool call monitoring.
pazdedavPersist agent progress to 00-session-state.json and resume from checkpoints.
Kling0012Optimize claude-flow v3 MCP server request handling with connection pooling and load balancing.
