Implement SkillSource.mjs for KB Ingestion

Context As part of Epic #11317, we need a new data source class to ingest skill markdown files into the Knowledge Base.

The Problem Skills are not currently semantically searchable because they lack a dedicated ingestion pipeline.

The Architectural Reality The Knowledge Base extraction pipeline uses a source-class pattern in ai/services/knowledge-base/source/. BaseSource is the parent class for all ingestion sources.

The Fix Create SkillSource.mjs extending BaseSource. It must extract .agents/skills/**/*.md files and attach the type: 'skill' chunk identifier along with skillName, sectionAnchor, triggerCondition, and isAtlasMonolithSubRule metadata.

Acceptance Criteria

• Create ai/services/knowledge-base/source/SkillSource.mjs extending BaseSource.
• Implement chunk extraction for .agents/skills/**/*.md files.
• Assign type: 'skill' to all generated chunks.
• Extract and attach skillName, sectionAnchor, triggerCondition, and isAtlasMonolithSubRule to the chunk metadata.

Out of Scope

• Integration into the main sync-kb run loop (handled in subsequent sub-ticket).
• Unit tests (handled in subsequent sub-ticket).

Origin Session ID: 2c4aa4df-2628-45ae-a9c2-156fd9308f21