Add a glossary of key terms to reduce ambiguity

[olaservo]

Background

Terms like "skill," "instruction," "context," "primitive," and "convention" are used in subtly different ways across documents and by different contributors. The distinction between "first-class skills" and "skills as context" (OQ #2) hinges entirely on definitional clarity. The convention draft (#43) will need shared terminology to be actionable.

This is also relevant for coordination with the Agent Skills spec, which has its own definitions that may or may not align with how this IG uses the same terms.

Scope

Create a glossary section — either in the README or as docs/glossary.md — defining at minimum:

1. Skill — How this IG defines it vs. how the Agent Skills spec defines it. Note any divergence.
2. Instruction — Server instructions (MCP concept) vs. skill instructions (workflow content). These are conflated in some discussions.
3. Primitive — MCP protocol primitive (tools, resources, prompts) vs. colloquial use.
4. Convention — A documented recommended pattern (not protocol spec) — important for framing Approach 6 and the v0.1 deliverable.
5. Context — As used in "skills as context" / "context-as-resources" — distinguish from the general sense of "LLM context window."
6. Progressive disclosure — Specific meaning in the skills context (lazy loading of skill content) vs. general UX usage.
7. Control model — Application-controlled vs. model-controlled resource access as defined by MCP, and how this applies to skills.
8. First-class primitive — What it means for something to be "first-class" in MCP (dedicated protocol methods, capability declaration, notifications).
9. Server author vs. skill author — These may be different people; some discussions assume they're the same.
10. Discovery — Finding that a skill exists vs. fetching its content. These are distinct operations that some discussions conflate.

Acceptance Criteria

• Glossary with at least 8 defined terms
• Each definition is 1–3 sentences, not exhaustive
• Notes alignment or divergence with Agent Skills spec where relevant
• Cross-referenced from README documents table
• Reviewed by at least 2 facilitators for accuracy

References

• Agent Skills Specification
• Open Question 2 — first-class vs. context
• Open Question 9 — control model
• Design Principles
• Convention draft: Draft MCP Skills Convention v0.1 #43