@zerwiz/ymir 0.1.4 → 0.1.5

package/.agents/tests/fm-harness-adapter-instructions-live-e2e.test.sh

@@ -0,0 +1,134 @@
+#!/usr/bin/env bash
+# Opt-in development evaluation for the harness-adapters routing instructions.
+# It sends the directly loaded router and a complete scenario set
+# to a local Ollama model, then compares the generated routing plan as normalized
+# JSON.
+# It makes no external-provider or remote CI call and does not claim that an absent or
+# unconfigured native harness loaded the references itself.
+set -u
+
+if [ "${FM_HARNESS_ADAPTER_INSTRUCTION_EVAL:-0}" != 1 ]; then
+  echo "skip: set FM_HARNESS_ADAPTER_INSTRUCTION_EVAL=1 and FM_HARNESS_ADAPTER_LOCAL_MODEL=<model> to run the local instruction evaluation"
+  exit 0
+fi
+
+# shellcheck source=tests/lib.sh
+. "$(dirname "${BASH_SOURCE[0]}")/lib.sh"
+
+ROUTER="$ROOT/.agents/skills/harness-adapters/SKILL.md"
+TMP_ROOT=$(fm_test_tmproot fm-harness-adapter-instructions)
+EXPECTED_JSON="$TMP_ROOT/expected.json"
+PROMPT_FILE="$TMP_ROOT/prompt.txt"
+RESPONSE_JSON="$TMP_ROOT/response.json"
+
+command -v curl >/dev/null 2>&1 || fail "curl is required for the local instruction evaluation"
+command -v jq >/dev/null 2>&1 || fail "jq is required for the local instruction evaluation"
+curl -fsS --max-time 2 http://127.0.0.1:11434/api/tags > "$TMP_ROOT/tags.json" \
+  || fail "local Ollama is unavailable at 127.0.0.1:11434; no remote provider fallback is allowed"
+
+MODEL=${FM_HARNESS_ADAPTER_LOCAL_MODEL:-}
+[ -n "$MODEL" ] || fail "FM_HARNESS_ADAPTER_LOCAL_MODEL must name an explicit local evaluator"
+jq -e --arg model "$MODEL" '.models | any(.name == $model)' "$TMP_ROOT/tags.json" >/dev/null \
+  || fail "requested local Ollama model is unavailable: $MODEL"
+
+cat > "$EXPECTED_JSON" <<'JSON'
+{
+  "cases": [
+    {"id":"start.default","common":["references/common/dispatch.md","references/common/model-and-effort.md"],"harness":"references/harness/claude.md"},
+    {"id":"start.trust-dialog","common":["references/common/control-and-recovery.md"],"harness":"references/harness/codex.md"},
+    {"id":"trust.default","common":["references/common/control-and-recovery.md"],"harness":"references/harness/opencode.md"},
+    {"id":"skill.default","common":["references/common/control-and-recovery.md"],"harness":"references/harness/pi.md"},
+    {"id":"interrupt.default","common":["references/common/control-and-recovery.md"],"harness":"references/harness/pi.md"},
+    {"id":"exit.default","common":["references/common/control-and-recovery.md"],"harness":"references/harness/grok.md"},
+    {"id":"resume.default","common":["references/common/control-and-recovery.md"],"harness":"references/harness/kimi.md"},
+    {"id":"recovery.default","common":["references/common/control-and-recovery.md"],"harness":"references/harness/cursor.md"},
+    {"id":"recovery.replacement-profile","common":["references/common/control-and-recovery.md","references/common/dispatch.md","references/common/model-and-effort.md"],"harness":"references/harness/muse.md"},
+    {"id":"recovery.secondmate","common":["references/common/control-and-recovery.md","references/common/primary-hooks.md"],"harness":"references/harness/claude.md"},
+    {"id":"recovery.replacement-secondmate","common":["references/common/control-and-recovery.md","references/common/dispatch.md","references/common/model-and-effort.md","references/common/primary-hooks.md"],"harness":"references/harness/codex.md"},
+    {"id":"primary.default","common":["references/common/primary-hooks.md"],"harness":"references/harness/opencode.md"},
+    {"id":"model-effort.default","common":["references/common/model-and-effort.md"],"harness":"references/harness/pi.md"},
+    {"id":"model-effort.configured-profile","common":["references/common/model-and-effort.md","references/common/dispatch.md"],"harness":"references/harness/pi.md"},
+    {"id":"verify.default","common":["references/common/dispatch.md","references/common/control-and-recovery.md","references/common/primary-hooks.md","references/common/model-and-effort.md"],"harness":"references/harness/grok.md"}
+  ]
+}
+JSON
+
+{
+  printf '%s\n' 'Act only as an evaluator of the directly loaded harness-adapters router below.'
+  printf '%s\n' 'For each requested operation.scenario, copy the common reference list in router order and append the requested harness reference in the harness field.'
+  printf '%s\n' 'Copy harness paths literally from the router map; never construct a filename from an identity, including when two identities share one path.'
+  printf '%s\n' 'The requests, in output order, are:'
+  printf '%s\n' \
+    'start.default claude' \
+    'start.trust-dialog codex' \
+    'trust.default opencode' \
+    'skill.default pi' \
+    'interrupt.default pi-signed' \
+    'exit.default grok' \
+    'resume.default kimi' \
+    'recovery.default cursor' \
+    'recovery.replacement-profile muse' \
+    'recovery.secondmate claude' \
+    'recovery.replacement-secondmate codex' \
+    'primary.default opencode' \
+    'model-effort.default pi' \
+    'model-effort.configured-profile pi-signed' \
+    'verify.default grok'
+  printf '%s\n' 'Return only one JSON object with a cases array; each item must have id, common, and harness fields.'
+  printf '%s\n' 'ROUTER START'
+  cat "$ROUTER"
+  printf '%s\n' 'ROUTER END'
+} > "$PROMPT_FILE"
+
+PAYLOAD=$(jq -n \
+  --arg model "$MODEL" \
+  --rawfile prompt "$PROMPT_FILE" \
+  '{model:$model,prompt:$prompt,stream:false,format:"json",options:{temperature:0,num_predict:4096}}')
+curl -fsS --max-time "${FM_HARNESS_ADAPTER_EVAL_TIMEOUT_SECONDS:-120}" \
+  -H 'Content-Type: application/json' \
+  -d "$PAYLOAD" http://127.0.0.1:11434/api/generate \
+  | jq -er '.response | fromjson' > "$RESPONSE_JSON" \
+  || fail "local model $MODEL did not return parseable routing JSON"
+jq '(.cases[]?.id) |= split(" ")[0]' "$RESPONSE_JSON" > "$TMP_ROOT/normalized-response.json" \
+  || fail "local model $MODEL returned an invalid routing case shape"
+
+if ! diff -u \
+  <(jq -S . "$EXPECTED_JSON") \
+  <(jq -S . "$TMP_ROOT/normalized-response.json") > "$TMP_ROOT/diff"; then
+  fail "local model $MODEL did not follow the routing instructions: $(tr '\n' ' ' < "$TMP_ROOT/diff")"
+fi
+pass "local model $MODEL selected every operation scenario and all nine harness identities"
+
+CHECKED=0
+MISSING=
+. "$ROOT/bin/fm-cursor-lib.sh"
+resolve_native_binary() {
+  local harness=$1 candidate
+  if [ "$harness" = cursor ]; then
+    fm_cursor_resolve_binary 2>/dev/null
+    return
+  fi
+  candidate=$(command -v "$harness" 2>/dev/null || true)
+  if [ -n "$candidate" ] && [ -x "$candidate" ]; then
+    printf '%s\n' "$candidate"
+    return 0
+  fi
+  if [ "$harness" = kimi ] && [ -n "${HOME:-}" ] && [ -x "$HOME/.kimi-code/bin/kimi" ]; then
+    printf '%s\n' "$HOME/.kimi-code/bin/kimi"
+    return 0
+  fi
+  return 1
+}
+
+for harness in claude codex opencode pi pi-signed grok kimi cursor muse; do
+  if binary=$(resolve_native_binary "$harness"); then
+    version=$("$binary" --version 2>/dev/null | head -1 | tr -d '\r') || version=unknown
+    printf '# native loader not claimed: %s %s is installed, but this harness-neutral evaluation does not exercise its provider transport\n' "$harness" "$version"
+    CHECKED=$((CHECKED + 1))
+  else
+    MISSING="$MISSING $harness"
+    printf '# unverified native loader: %s is not installed on this machine\n' "$harness"
+  fi
+done
+printf '# installed native tools recorded without overstating loader coverage: %s\n' "$CHECKED"
+[ -z "$MISSING" ] || printf '# unavailable native tools:%s\n' "$MISSING"

package/.agents/tests/fm-harness-adapter-references.test.sh

@@ -0,0 +1,30 @@
+#!/usr/bin/env bash
+# Portable structural validation for the harness-adapters routing artifact.
+set -u
+
+# shellcheck source=tests/lib.sh
+. "$(dirname "${BASH_SOURCE[0]}")/lib.sh"
+
+ROUTER="$ROOT/.agents/skills/harness-adapters/SKILL.md"
+TMP_ROOT=$(fm_test_tmproot fm-harness-adapter-references)
+ROUTING_JSON="$TMP_ROOT/routing.json"
+
+awk '
+  /^```json harness-adapter-routing-v1$/ { capture = 1; next }
+  capture && /^```$/ { exit }
+  capture { print }
+' "$ROUTER" > "$ROUTING_JSON"
+
+jq -e '
+  (.operations | type == "object") and
+  (.harnesses | type == "object") and
+  ([.operations[][] | select(type != "array")] | length == 0) and
+  ([.operations[][][] | select(type != "string")] | length == 0) and
+  ([.harnesses[] | select(type != "string")] | length == 0)
+' "$ROUTING_JSON" >/dev/null || fail "harness adapter routing artifact is not a normalized operation and harness map"
+
+jq -r '.operations[][][], .harnesses[]' "$ROUTING_JSON" | sort -u | while IFS= read -r path; do
+  [ -r "$ROOT/.agents/skills/harness-adapters/$path" ] \
+    || fail "harness adapter routing target is unreadable: $path"
+done
+pass "harness adapter routing artifact is normalized and every target is readable"

package/.agents/tests/fm-harness-liveness-drift-live-e2e.test.sh

@@ -0,0 +1,151 @@
+#!/usr/bin/env bash
+# tests/fm-harness-liveness-drift-live-e2e.test.sh - opt-in drift guard proving
+# every INSTALLED harness is still classified `alive` by the tmux liveness
+# probe (bin/backends/tmux.sh).
+#
+# Why this file exists: liveness classification depends on how a harness names
+# its own process, which is a surface the harness vendor controls and changes
+# without notice. Claude Code began reporting its version string as its process
+# name and became unattributable, which silently degraded supervision. A
+# regression that only a real harness release can cause needs a check that runs
+# real harnesses; a stubbed agent cannot see it, and neither can a table of
+# names transcribed from a previous release.
+#
+# Each harness is launched bare, with no prompt, so this consumes no model
+# tokens. The launch uses whatever credentials the harness already has; an
+# unauthenticated harness still starts its process, which is all the liveness
+# probe reads.
+#
+# Standard CI has no harness binaries or credentials, so this real-harness guard
+# is opt-in and on-demand. The portable counterpart in
+# tests/fm-tmux-agent-liveness.test.sh pins the classifier logic in CI. Run this
+# guard after any harness upgrade and before trusting refreshed evidence.
+set -u
+
+if [ "${FM_HARNESS_LIVENESS_DRIFT:-0}" != 1 ]; then
+  echo "skip: set FM_HARNESS_LIVENESS_DRIFT=1 to run the installed-harness liveness drift guard"
+  exit 0
+fi
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+
+fail() { printf 'not ok - %s\n' "$1" >&2; cleanup_all; exit 1; }
+pass() { printf 'ok - %s\n' "$1"; }
+note() { printf '# %s\n' "$1"; }
+
+command -v tmux >/dev/null 2>&1 || fail "tmux not found"
+REAL_TMUX=$(command -v tmux)
+SOCKET="fm-liveness-drift-$$"
+LAB=$(mktemp -d "${TMPDIR:-/tmp}/fm-liveness-drift.XXXXXX")
+SESSION=drift
+
+cleanup_all() {
+  "$REAL_TMUX" -L "$SOCKET" kill-server >/dev/null 2>&1 || true
+  [ -n "${LAB:-}" ] && rm -rf "$LAB"
+}
+trap cleanup_all EXIT
+
+mkdir -p "$LAB/shim" "$LAB/wt"
+cat > "$LAB/shim/tmux" <<SH
+#!/usr/bin/env bash
+exec "$REAL_TMUX" -L "$SOCKET" "\$@"
+SH
+chmod +x "$LAB/shim/tmux"
+PATH="$LAB/shim:$PATH"
+export PATH
+
+# shellcheck source=/dev/null
+. "$ROOT/bin/fm-backend.sh"
+# shellcheck source=/dev/null
+. "$ROOT/bin/fm-cursor-lib.sh"
+fm_backend_source tmux || fail "fm_backend_source tmux failed"
+
+"$REAL_TMUX" -L "$SOCKET" new-session -d -s "$SESSION" -n control -c "$LAB/wt" \
+  || fail "could not start the private tmux server"
+
+# Kimi is not required to be on PATH; mirror bin/fm-spawn.sh's own resolution
+# order so this guard covers the same binary firstmate would actually launch.
+resolve_harness_binary() {  # <harness>
+  local harness=$1 candidate
+  candidate=$(command -v "$harness" 2>/dev/null || true)
+  if [ -n "$candidate" ] && [ -x "$candidate" ]; then
+    printf '%s\n' "$candidate"
+    return 0
+  fi
+  if [ "$harness" = kimi ] && [ -n "${HOME:-}" ] && [ -x "$HOME/.kimi-code/bin/kimi" ]; then
+    printf '%s\n' "$HOME/.kimi-code/bin/kimi"
+    return 0
+  fi
+  # cursor is never on PATH under the name `cursor`: it installs as
+  # `cursor-agent` plus the legacy alias `agent`, and its user-local install is
+  # routinely absent from a non-interactive PATH. Resolve it through the same
+  # verified owner fm-spawn uses, so an unrelated executable named `agent` is
+  # rejected here exactly as it would be at launch.
+  if [ "$harness" = cursor ]; then
+    fm_cursor_resolve_binary 2>/dev/null && return 0
+    return 1
+  fi
+  return 1
+}
+
+CHECKED=0
+SKIPPED=
+
+# The verified adapters, in the order the harness-adapters skill router records
+# them. An adapter that gains a verified launch path belongs here too.
+# muse matters most of all here: its launcher execs a VERSION-SUFFIXED binary,
+# so the live process name changes on every auto-update and its install path
+# carries no `muse` component to fall back on. That is precisely the drift this
+# guard exists to catch, and only a real muse release can produce it.
+# cursor matters for the same reason muse does, from the other direction: it
+# runs as a bundled node script, so its pane title is a bare `node` that no name
+# pattern can own, and identity has to come from its install path or argv[0].
+for harness in claude codex opencode pi pi-signed grok kimi cursor muse; do
+  if ! bin_path=$(resolve_harness_binary "$harness"); then
+    SKIPPED="$SKIPPED $harness"
+    note "skip: $harness is not installed on this machine, so its classification is unverified here"
+    continue
+  fi
+
+  version=$("$bin_path" --version 2>/dev/null | head -1 | tr -d '\r') || version=
+  [ -n "$version" ] || version="unknown"
+
+  target="$SESSION:$harness"
+  # cursor blocks on a workspace-trust prompt in a directory it has never seen,
+  # which would hang this probe rather than classify anything; --trust is the
+  # same flag fm-spawn passes for the same reason.
+  launch_args=""
+  [ "$harness" = cursor ] && launch_args="--trust"
+  # shellcheck disable=SC2086  # deliberate: an empty value must add no argument
+  "$REAL_TMUX" -L "$SOCKET" new-window -d -t "$SESSION:" -n "$harness" -c "$LAB/wt" -- "$bin_path" $launch_args \
+    || fail "$harness ($version): could not launch a window for the liveness probe"
+
+  state=
+  for _ in $(seq 1 300); do
+    state=$(fm_backend_agent_state tmux "$target")
+    [ "$state" = alive ] && break
+    sleep 0.2
+  done
+
+  title=$(fm_backend_tmux_current_command "$target")
+  comms=$(fm_backend_tmux_foreground_comms "$target" | tr '\n' ' ')
+
+  [ "$state" = alive ] || fail \
+    "LIVENESS DRIFT: $harness $version is running but classifies '$state', not 'alive'. Supervision and lifecycle control treat this endpoint as unattributable. Observed process title '$title'; observed foreground process names [$comms]. Teach bin/backends/tmux.sh's fm_backend_tmux_classify_process_name the identity this release actually reports."
+
+  note "$harness $version: title='$title' foreground=[$comms]"
+
+  pass "harness liveness: $harness $version classifies alive"
+  CHECKED=$((CHECKED + 1))
+done
+
+[ "$CHECKED" -gt 0 ] || fail \
+  "no verified harness is installed here, so this run proved nothing; install at least one harness before trusting a pass"
+
+if [ -n "$SKIPPED" ]; then
+  note "unverified on this machine (not installed):$SKIPPED"
+fi
+note "checked $CHECKED installed harness(es)"
+
+cleanup_all
+trap - EXIT

package/.agents/tests/fm-herdr-lab.test.sh

@@ -0,0 +1,243 @@
+#!/usr/bin/env bash
+# Behavior tests for bin/fm-herdr-lab.sh using a stateful fake Herdr client.
+set -u
+
+# shellcheck source=tests/lib.sh
+. "$(dirname "${BASH_SOURCE[0]}")/lib.sh"
+
+TMP_ROOT=$(fm_test_tmproot fm-herdr-lab)
+FAKEBIN=$(fm_fakebin "$TMP_ROOT")
+FAKE_STATE="$TMP_ROOT/herdr-state"
+FAKE_LOG="$TMP_ROOT/herdr.log"
+TRIPWIRES="$TMP_ROOT/tripwires"
+REAL_SLEEP=$(command -v sleep)
+mkdir -p "$FAKE_STATE"
+printf '%s\n' '/home/test/.config/herdr/herdr.sock' > "$FAKE_STATE/default-socket"
+: > "$FAKE_LOG"
+
+cat > "$FAKEBIN/herdr" <<'SH'
+#!/usr/bin/env bash
+set -eu
+printf '%s\n' "$*" >> "$FM_FAKE_HERDR_LOG"
+state=$FM_FAKE_HERDR_STATE
+last=
+for arg in "$@"; do
+  previous=$last
+  last=$arg
+done
+[ "${previous:-}" = --session ] || { echo "fake herdr: missing trailing --session" >&2; exit 90; }
+session=$last
+default_socket=$(cat "$state/default-socket")
+lab_state=absent
+[ ! -f "$state/$session" ] || lab_state=$(cat "$state/$session")
+
+case "$1 ${2:-}" in
+  "session list")
+    if [ "$lab_state" = absent ] || [ "$lab_state" = deleted ]; then
+      jq -nc --arg socket "$default_socket" '{sessions:[{default:true,name:"default",running:true,socket_path:$socket}]}'
+    else
+      running=false
+      [ "$lab_state" = running ] && running=true
+      jq -nc --arg socket "$default_socket" --arg name "$session" --argjson running "$running" \
+        '{sessions:[{default:true,name:"default",running:true,socket_path:$socket},{default:false,name:$name,running:$running,socket_path:("/tmp/" + $name + ".sock")}]}'
+    fi
+    ;;
+  "server --session")
+    if [ "${FM_FAKE_HERDR_SERVER_DELAY:-0}" != 0 ]; then
+      "$FM_FAKE_HERDR_REAL_SLEEP" "$FM_FAKE_HERDR_SERVER_DELAY"
+    fi
+    printf '%s\n' running > "$state/$session"
+    ;;
+  "status --json")
+    if [ "$lab_state" = running ]; then
+      printf '%s\n' '{"server":{"running":true}}'
+    else
+      printf '%s\n' '{"server":{"running":false}}'
+    fi
+    ;;
+  "session stop")
+    [ "$3" = "$session" ] || exit 91
+    printf '%s\n' stopped > "$state/$session"
+    ;;
+  "session delete")
+    [ "$3" = "$session" ] || exit 92
+    [ "${FM_FAKE_HERDR_DELETE_FAIL:-}" != 1 ] || exit 93
+    printf '%s\n' deleted > "$state/$session"
+    ;;
+  *)
+    printf '%s\n' '{"ok":true}'
+    ;;
+esac
+SH
+chmod +x "$FAKEBIN/herdr"
+
+# shellcheck source=/dev/null
+. "$ROOT/bin/fm-herdr-lab.sh"
+
+run_with_fake() {
+  PATH="$FAKEBIN:$PATH" \
+    FM_FAKE_HERDR_STATE="$FAKE_STATE" \
+    FM_FAKE_HERDR_LOG="$FAKE_LOG" \
+    FM_FAKE_HERDR_REAL_SLEEP="$REAL_SLEEP" \
+    FM_FAKE_HERDR_SERVER_DELAY="${FM_FAKE_HERDR_SERVER_DELAY:-0}" \
+    FM_FAKE_HERDR_FAST_POLL="${FM_FAKE_HERDR_FAST_POLL:-}" \
+    FM_FAKE_HERDR_DELETE_FAIL="${FM_FAKE_HERDR_DELETE_FAIL:-}" \
+    FM_HERDR_LAB_STATE_DIR="$TRIPWIRES" \
+    "$@"
+}
+
+test_refuses_unsafe_names() {
+  local status=0 generated
+  fm_herdr_lab_validate_name default >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "literal default must be refused"
+  status=0
+  fm_herdr_lab_validate_name arbitrary-session >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "non-lab prefix must be refused"
+  fm_herdr_lab_validate_name fm-lab-safe-123 || fail "valid lab session name was refused"
+  generated=$(fm_herdr_lab_name fm-autodetect-smoke-concurrency-h3)
+  fm_herdr_lab_validate_name "$generated" || fail "generated lab session name was refused"
+  [ "${#generated}" -le 40 ] || fail "generated lab session name is too long for Herdr socket paths: $generated"
+  pass "fm-herdr-lab: names fail closed and require the lab prefix"
+}
+
+test_provision_run_and_guarded_teardown() {
+  local name='' line_count status=0 stop_line delete_line
+  name="fm-lab-behavior-$$"
+  : > "$FAKE_LOG"
+  run_with_fake fm_herdr_lab_provision "$name" || fail "provision failed"
+  [ "$(cat "$FAKE_STATE/$name")" = running ] || fail "provision did not start the named lab session"
+  assert_present "$TRIPWIRES/$name.fleet-state.json" "provision did not record the fleet-state tripwire"
+
+  run_with_fake fm_herdr_lab_cli "$name" workspace list >/dev/null || fail "safe run command failed"
+  run_with_fake fm_herdr_lab_cli "$name" server >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "bare server start outside provision must be refused"
+  status=0
+  run_with_fake fm_herdr_lab_cli "$name" server stop >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "server-global stop must be refused"
+  status=0
+  run_with_fake fm_herdr_lab_cli "$name" session delete "$name" >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "direct session delete must be refused"
+  status=0
+  run_with_fake fm_herdr_lab_cli "$name" status --session default >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "caller-supplied session flag must be refused"
+  status=0
+  run_with_fake fm_herdr_lab_cli "$name" status --session=default >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "caller-supplied equals-form session flag must be refused"
+  status=0
+  run_with_fake fm_herdr_lab_cli "$name" --handoff server stop >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "a leading option shifting server stop past the guard must be refused"
+  status=0
+  run_with_fake fm_herdr_lab_cli "$name" --no-session session delete "$name" >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "a leading option shifting session delete past the guard must be refused"
+  status=0
+  run_with_fake fm_herdr_lab_cli "$name" --remote host workspace list >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "a leading option subverting session isolation must be refused"
+
+  run_with_fake fm_herdr_lab_teardown "$name" || fail "guarded teardown failed"
+  [ "$(cat "$FAKE_STATE/$name")" = deleted ] || fail "teardown did not delete the lab session"
+  assert_absent "$TRIPWIRES/$name.fleet-state.json" "successful teardown left its tripwire behind"
+
+  while IFS= read -r line; do
+    case "$line" in
+      *"--session $name") : ;;
+      *) fail "Herdr call lacks a trailing lab session: $line" ;;
+    esac
+  done < "$FAKE_LOG"
+  line_count=$(wc -l < "$FAKE_LOG" | tr -d ' ')
+  stop_line=$(grep -n "^session stop $name --json --session $name$" "$FAKE_LOG" | cut -d: -f1)
+  delete_line=$(grep -n "^session delete $name --json --session $name$" "$FAKE_LOG" | cut -d: -f1)
+  if [ -z "$stop_line" ] || [ -z "$delete_line" ] || [ "$line_count" -le "$delete_line" ]; then
+    fail "teardown did not emit explicit stop/delete followed by the after tripwire"
+  fi
+  sed -n "$((stop_line - 1))p" "$FAKE_LOG" | grep -F "session list --json --session $name" >/dev/null \
+    || fail "stop was not immediately preceded by a fresh refuse-default session list"
+  sed -n "$((delete_line - 1))p" "$FAKE_LOG" | grep -F "session list --json --session $name" >/dev/null \
+    || fail "delete was not immediately preceded by a fresh refuse-default session list"
+  pass "fm-herdr-lab: provisioning, scoped calls, guarded teardown, and fleet tripwire are deterministic"
+}
+
+test_missing_tripwire_blocks_destruction() {
+  local name="fm-lab-no-tripwire-$$" status=0 before after
+  printf '%s\n' running > "$FAKE_STATE/$name"
+  : > "$FAKE_LOG"
+  before=$(wc -l < "$FAKE_LOG")
+  run_with_fake fm_herdr_lab_teardown "$name" >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "missing tripwire must refuse teardown"
+  after=$(wc -l < "$FAKE_LOG")
+  [ "$before" = "$after" ] || fail "missing tripwire reached Herdr instead of refusing before destructive calls"
+  pass "fm-herdr-lab: missing tripwire refuses teardown before any Herdr call"
+}
+
+test_changed_default_trips_after_teardown() {
+  local name="fm-lab-tripwire-change-$$" status=0
+  : > "$FAKE_LOG"
+  run_with_fake fm_herdr_lab_provision "$name" || fail "tripwire fixture provision failed"
+  printf '%s\n' '/changed/default.sock' > "$FAKE_STATE/default-socket"
+  run_with_fake fm_herdr_lab_teardown "$name" >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "changed default fleet state must fail teardown"
+  assert_present "$TRIPWIRES/$name.fleet-state.json" "failed tripwire should retain evidence"
+  printf '%s\n' '/home/test/.config/herdr/herdr.sock' > "$FAKE_STATE/default-socket"
+  rm -f "$TRIPWIRES/$name.fleet-state.json"
+  pass "fm-herdr-lab: changed default fleet state is a hard failure"
+}
+
+test_stopped_owned_lab_can_reprovision() {
+  local name="fm-lab-reprovision-$$"
+  : > "$FAKE_LOG"
+  run_with_fake fm_herdr_lab_provision "$name" || fail "initial provision failed"
+  run_with_fake fm_herdr_lab_stop "$name" || fail "guarded stop failed"
+  [ "$(cat "$FAKE_STATE/$name")" = stopped ] || fail "guarded stop did not stop the lab session"
+  assert_present "$TRIPWIRES/$name.fleet-state.json" "stop removed the lab ownership tripwire"
+  run_with_fake fm_herdr_lab_provision "$name" || fail "re-provision after guarded stop failed"
+  [ "$(cat "$FAKE_STATE/$name")" = running ] || fail "re-provision did not restart the stopped lab session"
+  assert_present "$TRIPWIRES/$name.fleet-state.json" "re-provision removed the lab ownership tripwire"
+  run_with_fake fm_herdr_lab_teardown "$name" || fail "teardown after re-provision failed"
+  pass "fm-herdr-lab: an owned stopped lab can re-provision safely"
+}
+
+test_failed_delete_retains_tripwire() {
+  local name="fm-lab-delete-failure-$$" status=0
+  : > "$FAKE_LOG"
+  run_with_fake fm_herdr_lab_provision "$name" || fail "delete-failure fixture provision failed"
+  FM_FAKE_HERDR_DELETE_FAIL=1 run_with_fake fm_herdr_lab_teardown "$name" >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "failed delete must fail teardown"
+  [ "$(cat "$FAKE_STATE/$name")" = stopped ] || fail "failed delete unexpectedly removed the lab session"
+  assert_present "$TRIPWIRES/$name.fleet-state.json" "failed delete removed the ownership tripwire"
+  run_with_fake fm_herdr_lab_teardown "$name" || fail "retry after failed delete did not clean up the lab session"
+  assert_absent "$TRIPWIRES/$name.fleet-state.json" "successful retry left the ownership tripwire behind"
+  pass "fm-herdr-lab: failed deletion retains ownership until absence is confirmed"
+}
+
+test_timed_out_provision_cancels_late_launch() {
+  local name="fm-lab-late-launch-$$" status=0
+  cat > "$FAKEBIN/sleep" <<'SH'
+#!/usr/bin/env bash
+if [ "${FM_FAKE_HERDR_FAST_POLL:-}" = 1 ]; then
+  exit 0
+fi
+exec "$FM_FAKE_HERDR_REAL_SLEEP" "$@"
+SH
+  chmod +x "$FAKEBIN/sleep"
+  : > "$FAKE_LOG"
+  FM_FAKE_HERDR_FAST_POLL=1 FM_FAKE_HERDR_SERVER_DELAY=30 \
+    run_with_fake fm_herdr_lab_provision "$name" >/dev/null 2>&1 || status=$?
+  expect_code 1 "$status" "timed-out provision must fail"
+  assert_present "$TRIPWIRES/$name.fleet-state.json" \
+    "timed-out provision must retain its tripwire until teardown"
+  run_with_fake fm_herdr_lab_teardown "$name" || fail "teardown after timed-out provision failed"
+  assert_absent "$TRIPWIRES/$name.fleet-state.json" \
+    "teardown after timed-out provision did not remove its tripwire"
+  "$REAL_SLEEP" 1.1
+  if [ -f "$FAKE_STATE/$name" ] && [ "$(cat "$FAKE_STATE/$name")" = running ]; then
+    fail "timed-out provision left a late-starting lab session after teardown"
+  fi
+  pass "fm-herdr-lab: timed-out provisioning cancels the launch before teardown"
+}
+
+test_refuses_unsafe_names
+test_provision_run_and_guarded_teardown
+test_missing_tripwire_blocks_destruction
+test_changed_default_trips_after_teardown
+test_stopped_owned_lab_can_reprovision
+test_failed_delete_retains_tripwire
+test_timed_out_provision_cancels_late_launch