Implements AgentSkillDiscovery class supporting three-tier progressive disclosure model:
- Global scope: organization-level skill directories (lowest precedence)
- Project scope: project-level skill directories (intermediate precedence)
- Local scope: working-directory skill directories (highest precedence)

Key changes:
- Added agent_skill_discovery.py with AgentSkillDiscovery, Scope, ScopeOrder, DiscoveredScopedSkill, MultiScopeDiscoveryResult, ScopedDiscoveryConflict types
- Updated skills module __init__.py to export new discovery classes
- Added BDD test file agent_skills_multi_scope_discovery.feature covering all scope scenarios
- Added step definitions agent_skills_multi_scope_discovery_steps.py with 30+ scenarios
- Updated CHANGELOG.md under [Unreleased] section
- Updated CONTRIBUTORS.md

Name collision resolution: local > project > global (lower ScopeOrder value wins)

ISSUES CLOSED: #9369

@@ -0,0 +150,4 @@
+    context._discovery_result = None
+
+    try:
+        discovery = AgentSkillDiscovery(

@@ -0,0 +366,4 @@
+    """
+    registry = getattr(context, "_scoped_dir_registry", {})
+    context._discovery_errors = None
+    context._scop_result = None

@@ -0,0 +233,4 @@
+                existing_precedence = _scope_precedence(scope)
+                winner_precedence = _scope_precedence(existing.scope)
+
+                if existing_precedence < winner_precedence:

@@ -0,0 +268,4 @@
+
+def _scope_precedence(scope: Scope) -> int:
+    """Return the numeric precedence of a scope (lower = higher priority)."""
+    return ScopeOrder[scope.name.upper()].value  # type: ignore[misc]

@@ -0,0 +426,4 @@
+        # Initialize all scopes so the dict is complete
+        for s in Scope:
+            if s not in result.scoped_skills:
+                result.scoped_skills[s] = []  # type: ignore[index]