graph.json never populated with extracted wikilinks

[tashisleepy]

Context

Tracked from issue #3 review as deferred item.

Current behavior

schema/graph.json contains 9 nodes and 19 edges, all populated manually. The wiki markdown files use [[wikilinks]] syntax for cross-references (per .claude/rules/knowledge-engine.md), but nothing extracts these at ingest time to populate graph.json.

Expected behavior

When a wiki page is added or updated (via save-session.sh or similar), the system should:

1. Parse the markdown for [[wikilink]] patterns
2. Resolve each wikilink to a target node ID
3. Add edges to schema/graph.json with source/target/type
4. Handle broken wikilinks (target doesn't exist) - flag or skip

Acceptance criteria

• Wikilink extractor function added (probably in bridge.py or new graph_builder.py)
• Runs automatically on save-session.sh and any other ingest path
• Populates schema/graph.json with deduplicated edges
• Broken wikilinks logged to log.md for manual review
• Existing manual edges preserved, not overwritten

Priority

Medium - affects graph-based querying and wiki health checks, but manual edges still work.

Related

• Deferred from Security + correctness findings from independent review #3 security review
• Rule source: .claude/rules/knowledge-engine.md (wikilink syntax requirement)