const modified: string[] = [];
for (const [id, node1] of Array.from(nodes1.entries())) {
    const node2 = nodes2.get(id);
    if (node2 && node1.data !== node2.data) {
        modified.push(id);
    }
}
Node IDs are computed as SHA-256(data). Since the ID is the hash of the data, it is cryptographically impossible for two nodes to share the same ID but have different data. The condition node2 && node1.data !== node2.data can never be true.

When a file is modified, its content hash changes, which changes its node data ("path:hash"), which changes its node ID. The old node ID appears in removed and the new one appears in added — but modified is always [].

Why the system still works correctly
MerkleDAG.compare() is only used as a fast "did anything change?" boolean gate inside FileSynchronizer.checkForChanges(). The actual categorization of files into added/removed/modified is done by compareStates(), which compares Map<filePath, contentHash> — keyed by file path, not content hash — so it correctly identifies modifications.

MerkleDAG.compare()     →  "something changed" (boolean gate)
    ↓
compareStates()         →  "here's exactly what changed" (added/removed/modified)
Changes
merkle.ts: Removed the unreachable modified loop from compare(). Simplified the return type from { added, removed, modified } to { added, removed }. Replaced Map with Set since only node IDs are needed for the comparison.
synchronizer.ts: Removed changes.modified check from the DAG comparison condition to match the updated return type.
Verification
pnpm build:core — passes
pnpm build:mcp — passes
No behavioral change: file modification detection continues to work via compareStates() in FileSynchronizer