learn-from-skill

name	learn-from-skill
description	Retrieves and presents agentskills.io skill content as structured guidance. Use when the user asks to "learn about", "explain how to", "what's the best practice for", "guide me through", "teach me", or wants knowledge from a skill without executing commands. Reads the skill and presents its guidance in a conversational format.
metadata	{"author":"Troy Taylor","version":"1.0","pattern":"aggregation"}

Learn From a Skill

What This Skill Does

Loads an agentskills.io skill and presents its guidance — decision frameworks, best practices, code patterns, validation checklists — in a conversational format. No code execution needed. The skill content enriches the response with curated expert knowledge.

When to Activate

• User asks "how should I" or "what's the best approach for"
• User asks to "explain" or "teach me" a topic
• User says "load that skill" after a discovery step
• User wants guidance/knowledge without running commands
• User asks about architecture decisions, patterns, or best practices

Workflow

1. Identify which skill to load. Either from a prior discovery step or by searching:

   search_skills(query: "AI technology selection .NET")

2. Load the full skill content. Use get_skill to fetch the SKILL.md:

   get_skill(owner: "dotnet", repo: "skills", skill_path: "plugins/dotnet-ai/skills/technology-selection")

3. Read and internalize the skill. Parse the returned content for:

   • Decision trees or frameworks
   • Step-by-step workflows
   • Code examples and patterns
   • Validation checklists
   • Anti-patterns to avoid

4. Respond using the skill's guidance. Don't just dump the raw content. Apply it to the user's specific question:

   • If they asked "which AI library should I use?" → walk them through the skill's decision tree with their specific inputs
   • If they asked "how do I design a database?" → present the schema design workflow adapted to their scenario

5. Cite the source. Include a link to the skill on GitHub so the user can bookmark it or explore further.

Output Format

For decision guidance

Based on the technology-selection skill:

For your scenario (document classification on structured data):

→ Use ML.NET — not an LLM.

Your requirement	Recommendation	Reason
Structured tabular data	ML.NET	Faster, cheaper, deterministic
Classification task	ML.NET AutoML	Purpose-built for this
.NET 8+ project	Compatible	Full support

Key guardrails from the skill:

• Always set MLContext(seed: 42) for reproducibility
• Use TrainTestSplit — never evaluate on training data
• Start with AutoML, then refine manually

Source: dotnet/skills — technology-selection

For process guidance

Based on the requirements-clarity skill:

Before implementing, ask these two questions:

1. Why? (YAGNI check) — Is this actually needed?
2. Simpler? (KISS check) — Is there a simpler approach?

[continued workflow steps from the skill...]

Handling Edge Cases

• Skill not found: Suggest searching with different keywords or browsing the curated registry.
• Skill too long: Summarize the key sections and offer to deep-dive into any specific part.
• Skill requires execution: Note that the skill has executable steps and offer to switch to apply-skill mode.