@team-agent/installer 0.4.2 → 0.4.3

package/Cargo.lock

@@ -566,7 +566,7 @@ dependencies = [
 
 [[package]]
 name = "team-agent"
-version = "0.4.2"
+version = "0.4.3"
 dependencies = [
  "anyhow",
  "chrono",

package/Cargo.toml

@@ -9,7 +9,7 @@ members = ["crates/team-agent"]
 
 [workspace.package]
 edition = "2021"
-version = "0.4.2"
+version = "0.4.3"
 license = "AGPL-3.0"
 rust-version = "1.95"
 

package/crates/team-agent/src/lifecycle/launch.rs

@@ -4850,6 +4850,16 @@ fn spec_agents(spec: &Value) -> Vec<AgentId> {
         .collect()
 }
 
+/// Bug 1 (0.4.2): expose spec agent id set so the restart path can filter
+/// state.agents to only the agents currently defined in the rebuilt spec.
+/// Returns a `BTreeSet<String>` for O(log n) membership checks.
+pub(crate) fn spec_agent_id_set(spec: &Value) -> std::collections::BTreeSet<String> {
+    spec_agent_values(spec)
+        .into_iter()
+        .filter_map(|agent| agent.get("id").and_then(Value::as_str).map(str::to_string))
+        .collect()
+}
+
 fn spec_agent_values(spec: &Value) -> Vec<&Value> {
     spec.get("agents")
         .and_then(Value::as_list)

package/crates/team-agent/src/lifecycle/restart/rebuild.rs

@@ -142,6 +142,41 @@ pub fn restart_with_transport_with_session_convergence_deadline(
         LifecycleError::TeamSelect("active team spec workspace not found".to_string())
     })?;
     let safety = crate::lifecycle::launch::effective_runtime_config(&spec)?;
+    // Bug 1 (0.4.2 P0): the rebuilt spec is the single source of truth for the
+    // active roster. Any agent in state.agents that is NOT in the rebuilt
+    // spec is a stale leftover (role doc was deleted between sessions) and
+    // must NOT be restarted — otherwise a `team-agent restart` would re-spawn
+    // a removed worker from stale state. Prune state.agents in-place before
+    // convergence/plan/spawn so every downstream step (resume validation,
+    // event log, spawn loop) sees the bounded roster.
+    let spec_agent_ids = crate::lifecycle::launch::spec_agent_id_set(&spec);
+    if let Some(agents_obj) = state.get_mut("agents").and_then(|v| v.as_object_mut()) {
+        let stale_ids: Vec<String> = agents_obj
+            .keys()
+            .filter(|id| !spec_agent_ids.contains(id.as_str()))
+            .cloned()
+            .collect();
+        for stale_id in &stale_ids {
+            agents_obj.remove(stale_id);
+            let _ = crate::event_log::EventLog::new(&selected.run_workspace).write(
+                "restart.agent_skipped_not_in_spec",
+                serde_json::json!({
+                    "agent_id": stale_id,
+                    "team_key": selected.team_key.as_str(),
+                    "reason": "agent present in state.agents but absent from rebuilt spec; \
+                               role doc was removed — restart will not respawn removed workers",
+                    "action": format!(
+                        "to permanently remove run `team-agent remove-agent {stale_id}`; \
+                         to re-add restore the role doc under agents/<id>.md and run \
+                         `team-agent restart`"
+                    ),
+                }),
+            );
+        }
+        if !stale_ids.is_empty() {
+            save_restart_state(&selected.run_workspace, &mut state, &selected.team_key)?;
+        }
+    }
     let mut convergence = converge_missing_provider_sessions(
         &mut state,
         session_convergence_deadline(session_converge_deadline_ms),

package/crates/team-agent/src/provider/adapter.rs

@@ -1072,6 +1072,12 @@ fn scan_session_candidates_once(
         // identity match (TEAM_AGENT_ID literal in the transcript head, or
         // path-encoded agent_id). If none, return empty — capture stays
         // pending/ambiguous, which is safer than misattribution.
+        //
+        // Bug 2 (0.4.2 P0) reinforcement: even when positive-identity
+        // filtering returns no candidates, do NOT let downstream
+        // time-window narrowing run for Claude when expected_session_id
+        // misses — that's exactly the leader-session-leak symptom the
+        // architect cataloged in .team/artifacts/bug-042-restart-scope-and-session.md.
         if matches!(provider, Provider::Claude | Provider::ClaudeCode) {
             let positive_only: Vec<CapturedSessionCandidate> = out
                 .iter()
@@ -2044,6 +2050,50 @@ mod e6_session_attribution_tests {
         f.set_modified(when).unwrap();
     }
 
+    /// Bug 2 (0.4.2 P0): when expected_session_id is set (restart --allow-fresh
+    /// pre-allocates a UUID via --session-id) but the Claude CLI didn't adopt
+    /// it (no <expected>.jsonl on disk), capture must NOT fall back to
+    /// time-window narrowing over leader transcripts sharing the same cwd.
+    /// Pre-fix: leader transcript with later mtime would be silently picked
+    /// up as the worker session. Post-fix: empty list returned (caller treats
+    /// as session-not-yet-captured and retries on next tick).
+    #[test]
+    fn scan_expected_session_id_miss_refuses_to_pick_leader_sibling() {
+        let base = tmp_root("strict-no-leader-fallback");
+        let cwd = base.join("ws");
+        std::fs::create_dir_all(&cwd).unwrap();
+        let proj = base.join("projects");
+        // Simulate the leader's transcript and a stale earlier transcript in
+        // the same cwd. NEITHER matches expected_session_id.
+        let leader = write_transcript(&proj, "11111111-1111-4111-8111-111111111111", &cwd);
+        let stale = write_transcript(&proj, "22222222-2222-4222-8222-222222222222", &cwd);
+        let _ = (leader, stale);
+        let ctx = CaptureSessionContext {
+            agent_id: "claude-worker".to_string(),
+            spawn_cwd: cwd.clone(),
+            pane_id: None,
+            pane_pid: None,
+            spawned_at: Some("2020-01-01T00:00:00+00:00".to_string()),
+            // Worker was spawned with --session-id <expected> but Claude
+            // didn't write <expected>.jsonl.
+            expected_session_id: Some(SessionId::new(
+                "99999999-9999-4999-8999-999999999999",
+            )),
+            provider_projects_root: Some(proj.clone()),
+        };
+        let out = scan_session_candidates_once(Provider::ClaudeCode, &ctx).unwrap();
+        assert!(
+            out.is_empty(),
+            "expected_session_id set + no exact match → must NOT fall back to \
+             time-window narrowing (would grab leader's transcript). \
+             got={:?}",
+            out.iter()
+                .filter_map(|c| c.captured.session_id.as_ref().map(|s| s.as_str().to_string()))
+                .collect::<Vec<_>>()
+        );
+        let _ = std::fs::remove_dir_all(&base);
+    }
+
     // ── E11 层1:copilot session 归因(expected-id 优先,不抓 leader latest)──
     struct HomeGuard {
         prev: Option<std::ffi::OsString>,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@team-agent/installer",
-  "version": "0.4.2",
+  "version": "0.4.3",
   "description": "npx installer for Team Agent",
   "keywords": [
     "codex",
@@ -20,9 +20,9 @@
     "team-agent-installer": "npm/install.mjs"
   },
   "optionalDependencies": {
-    "@team-agent/cli-darwin-arm64": "0.4.2",
-    "@team-agent/cli-darwin-x64": "0.4.2",
-    "@team-agent/cli-linux-x64": "0.4.2"
+    "@team-agent/cli-darwin-arm64": "0.4.3",
+    "@team-agent/cli-darwin-x64": "0.4.3",
+    "@team-agent/cli-linux-x64": "0.4.3"
   },
   "scripts": {
     "postinstall": "node npm/bincheck.mjs",