test: gate st2 against bounded pty fleet snapshots

agent-session-id: dev3.dotfiles-cos-misc-agent-runtime-simplification
agent-tool: Codex
agent-tool-version: 0.145.0
agent-model: gpt-5.6-sol
agent-runtime-profile: /home/schickling/.config/coding-agents/profile.json
agent-skills-manifest: /nix/store/nk9iml2841l1yjjg0f6f0d3y60zkg1nn-agent-skills-corpus/share/agent-skills/manifest.json
tooling-profile: dotfiles@4a0515f

Author: schickling-assistant

flake.lock

@@ -4,13 +4,18 @@
   inputs = {
     nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
     flake-utils.url = "github:numtide/flake-utils";
+    # Runtime compatibility gate: this is the smallest pty revision containing both the
+    # ambiguity-safe PID read and the single fleet-wide socket fallback budget.
+    pty.url = "github:compoundingtech/pty/6de78841e05ac5b9f4fbe844a71b6dfcc7c7f9a5";
+    pty.inputs.nixpkgs.follows = "nixpkgs";
   };
 
   outputs =
     {
       self,
       nixpkgs,
       flake-utils,
+      pty,
     }:
     flake-utils.lib.eachDefaultSystem (
       system:
@@ -151,6 +156,43 @@
         # wants them.
         checks.st2 = st2;
 
+        # Real producer-consumer contract: st2 consumes `pty list --json` from the exact pty
+        # revision that owns fleet observation. Fake CLI fixtures below still cover malformed
+        # output and a wedged child; this check proves the healthy 0/75/100/500-seat path crosses
+        # both packaged binaries within st2's short outer deadline.
+        checks.pty-fleet-contract = pkgs.runCommand "st2-pty-fleet-contract-${version}" {
+          nativeBuildInputs = [
+            pkgs.coreutils
+            pty.packages.${system}.default
+            st2
+          ];
+        } ''
+          export HOME=$(mktemp -d)
+          catalog=$(mktemp -d)
+          mkdir -p "$catalog/agents/contract/gone"
+          printf '%s\n' \
+            'agent "gone" { host "contract"; retired #true; command "true" }' \
+            > "$catalog/agents/contract/gone/agent.kdl"
+
+          for fleet_size in 0 75 100 500; do
+            root=$(mktemp -d)
+            i=0
+            while test "$i" -lt "$fleet_size"; do
+              seat=$(printf 'seat-%03d' "$i")
+              : > "$root/$seat.sock"
+              printf '%s\n' "$$" > "$root/$seat.pid"
+              i=$((i + 1))
+            done
+
+            PTY_ROOT="$root" timeout 2s \
+              st2 doctor --catalog "$catalog" --host contract \
+              > "doctor-$fleet_size.out"
+            grep -F 'contract.gone retirement complete' "doctor-$fleet_size.out" >/dev/null
+          done
+
+          touch $out
+        '';
+
         # Smoke test that the built binary actually runs and its command tree is
         # wired, independent of the in-tree `cargo test`.
         checks.help = pkgs.runCommand "st2-help-${version}" { } ''

flake.nix

@@ -188,6 +188,36 @@ fn doctor_bounds_a_hung_pty_probe_and_reports_the_runtime_error() {
     assert!(stdout.contains("timed out after 2.0s"), "{stdout}");
 }
 
+#[test]
+fn doctor_rejects_a_partial_pty_snapshot_atomically() {
+    let tmp = tempfile::tempdir().unwrap();
+    let catalog = tmp.path().join("catalog");
+    let declaration = catalog.join("agents/h/gone/agent.kdl");
+    let bin = tmp.path().join("bin");
+    fs::create_dir_all(declaration.parent().unwrap()).unwrap();
+    fs::create_dir_all(&bin).unwrap();
+    fs::write(
+        declaration,
+        "agent \"gone\" { host \"h\"; retired #true; command \"true\" }\n",
+    )
+    .unwrap();
+    executable(
+        &bin.join("pty"),
+        "#!/bin/sh\nprintf '[{\"name\":\"h.gone.agent\",\"status\":\"running\"},'\n",
+    );
+
+    let output = doctor(&catalog, &bin, &tmp.path().join("state"));
+
+    assert!(!output.status.success());
+    let stdout = String::from_utf8_lossy(&output.stdout);
+    assert!(stdout.contains("✗ task runtime readable"), "{stdout}");
+    assert!(stdout.contains("parsing `pty list --json`"), "{stdout}");
+    assert!(
+        !stdout.contains("retirement complete"),
+        "a valid prefix must never be consumed as a partial snapshot:\n{stdout}"
+    );
+}
+
 #[test]
 fn retired_declaration_is_healthy_when_tasks_and_presence_are_absent() {
     let tmp = tempfile::tempdir().unwrap();