refactor(uuid): resolve B-10 by deprecating agent_identities and removing PyYAML dependency

- Remove dead agent_identities read path and PyYAML import from workspace_uuid.py
- Defer eager PyYAML import in verify_session.py to lazy YAML fallback branch
- Simplify UUID resolution to 2-tier model (tier-1 own row ID -> tier-2 adapter scan)
- Clean up unused Drift D and ghost cache clearing in reconcile.sh and stop_session.sh
- Add 3 regression guards in tests/test_tier1_unit.py (266/266 PASS)
- Update IMPROVEMENTS.md, VERSIONS.md, and SKILL.md files

.agents/skills/multi-agent-mux-resume/SKILL.md

@@ -47,15 +47,12 @@ ideal resume path:
 
 ## UUID resolution order
 
-`agent-sessions.yaml` is the *primary* source. The skill reads in this order:
+`agent-sessions.yaml` and on-disk discovery are used to resolve the UUID in this order:
 
-1. **`agent-sessions.yaml` → `agent_identities.<agent>.session_id` (claude) / `conversation_id` (agy)** — explicit saved value
-2. **`agent-sessions.yaml` → `agent_identities.<agent>.session_jsonl` (claude) / `conversation_db` (agy)** — the on-disk artifact
-3. **Fallback: scan disk for the workspace's most recent conversation** (Note: `CLAUDE_PROJECT_DIR` overrides the default `~/.claude/projects/` path, and `HOME_DIR` overrides the `~` path) —
-   - claude: `ls -t $CLAUDE_PROJECT_DIR/<workspace-key>/*.jsonl | head -1` and parse the `sessionId` from the first line
-   - agy: `jq -r '."<workspace>"' $HOME_DIR/.gemini/antigravity-cli/cache/last_conversations.json`
+1. **`herdr_sessions[]` row's per-row own id** (`claude_session_id_own` / `agy_conversation_id_own` / `hermes_conversation_id_own` / `cline_conversation_id_own`) — explicitly saved by `multi-agent-mux-stop` right before teardown (tier-1, race-free).
+2. **Workspace-scoped on-disk scan** (adapter `discover()`)
 
-If all three are empty → the workspace has no conversation yet. Fall back to `multi-agent-mux-create`.
+If both are empty → the workspace has no conversation yet. Fall back to `multi-agent-mux-create`.
 
 ## Workflow