scanGemini() in bin/gstack-global-discover.ts (around line 372) scans ~/.gemini/tmp/<project>/chats/session-*.json — the legacy Gemini CLI JSON format. The Antigravity CLI (agy), which shares the ~/.gemini/ home directory, stores its sessions as SQLite under ~/.gemini/antigravity-cli/conversations/*.db. These are never scanned, so all agy usage is invisible to discovery.

Evidence (my machine, 2026-06-11)

• ~/.gemini/antigravity-cli/conversations/: 504 .db files, newest mtime today (2026-06-11 23:44).
• ~/.gemini/tmp/**/session-*.json modified in the last 14 days: 0.
• Net effect: the gemini tool count in discover output is effectively always ~0 for any agy-primary user, silently undercounting cross-tool activity.

Why it isn't a one-line path swap

The agy .db schema is blob-encoded — tables are trajectory_meta, steps, gen_metadata, executor_metadata, parent_references, trajectory_metadata_blob, battle_mode_infos — with no plain cwd/path/repo column. discover's repo-mapping relies on cwd, which here lives inside a protobuf/JSON blob payload. Supporting agy therefore means a new SQLite scanner that opens each .db and decodes a blob to recover cwd (perf note: hundreds of dbs per machine).

Suggested fix

Add a scanAntigravity() alongside scanGemini(), gated on existsSync(~/.gemini/antigravity-cli/conversations), surfaced either under the existing gemini bucket or a new agy tool tag.

Reported from gstack 1.57.10.0.