Agent Lifecycle Commands (Phase 8)

Phase 8 adds a set of operator-facing lifecycle commands. All commands target an AI lenser by its handle (e.g. @my-lenser).

---

Command overview

Command	Description	RPC called
lf kill-switch on <handle>	Activate the kill switch — denies all new runs	fn_set_kill_switch
lf kill-switch off <handle>	Deactivate the kill switch	fn_set_kill_switch
lf kill-switch status <handle>	Show current kill switch state	fn_evaluate_pre_run_policy
lf lenser ai pause <handle>	Pause new run dispatch	fn_set_runner_paused
lf lenser ai resume <handle>	Resume run dispatch	fn_set_runner_paused
lf lenser ai status <handle>	Show lenser state and active run count	—
lf run cancel <run-id>	Cancel a running or queued run	fn_cancel_team_run
lf run report <run-id>	Print the run report for a terminal run	—
lf run incidents <run-id>	List incidents for a run	—
lf run policy-check <handle>	Run a pre-run policy check and show verdict	fn_evaluate_pre_run_policy
lf budget set <handle>	Set or update the budget ceiling	fn_set_workspace_setting
lf budget status <handle>	Show current spend and ceiling	—
lf dark-launch enable <handle>	Enable dark launch for a workflow	fn_set_dark_launch
lf dark-launch disable <handle>	Disable dark launch	fn_set_dark_launch
lf dark-launch status <handle>	Show dark launch percentage and included workflow IDs	—
lf policy log <handle>	Stream recent policy evaluation rows	—
lf policy stats <handle>	Aggregate policy verdict counts for a period	—

---

Kill Switch

Usage

lf kill-switch on|off|status <handle>

Arguments

Arg	Description
on	Activate the kill switch
off	Deactivate the kill switch
status	Print current state
<handle>	AI lenser handle (e.g. @my-lenser)

Flags

Flag	Description
--reason <text>	Optional reason stored in policy_evaluations.reason
--workspace <id>	Target a specific workspace (defaults to active session workspace)

Example

lf kill-switch on @my-lenser --reason "suspected prompt injection"

Kill switch activated for @my-lenser
  reason: suspected prompt injection
  active runs will be marked killed on next heartbeat

lf kill-switch status @my-lenser

Kill switch status for @my-lenser
  state:    ACTIVE
  set at:   2026-05-01T14:22:00Z
  reason:   suspected prompt injection

RPC called: fn_set_kill_switch(p_ai_lenser_id, p_active, p_reason)

---

Pause / Resume

Usage

lf lenser ai pause|resume|status <handle>

Arguments

Arg	Description
pause	Pause new run dispatch; active runs complete
resume	Re-enable dispatch
status	Print lenser state and active run count
<handle>	AI lenser handle

Flags

Flag	Description
--workspace <id>	Target workspace

Example

lf lenser ai pause @my-lenser

Lenser paused for @my-lenser
  active runs (2) will complete normally
  new runs will be denied with verdict: pause

lf lenser ai status @my-lenser

Lenser status for @my-lenser
  state:        PAUSED
  active runs:  2
  queued runs:  0

RPC called: fn_set_runner_paused(p_ai_lenser_id, p_paused)

---

Run Control

Usage

lf run cancel|report|incidents|policy-check <id-or-handle>

lf run cancel

Cancel an active or queued run:

lf run cancel run_abc123

Run run_abc123 cancelled
  a run_report will be created with outcome: cancelled

RPC called: fn_cancel_team_run(p_team_run_id)

lf run report

Print the run report for a terminal run:

lf run report run_abc123

Run Report  run_abc123
──────────────────────────────────
  outcome             success
  steps               12
  tool invocations    3
  memory writes       2
  cost estimate       $0.04
  duration            4 231 ms
  summary             Completed sentiment analysis across 10 inputs.

lf run incidents

List incidents attached to a run report:

lf run incidents run_xyz789

Incidents for run_xyz789
  [high]    budget_overrun     Cumulative cost $12.80 exceeded ceiling $10.00
  [medium]  tool_timeout       Tool search_web timed out after 30s on step 4

lf run policy-check

Run a pre-run policy check without dispatching:

lf run policy-check @my-lenser --workflow wf_abc123

Policy check for @my-lenser, workflow wf_abc123
  ✓ kill_switch    not active
  ✓ runner_pause   not paused
  ✓ budget_ceiling $12.40 / $100.00
  ✓ parallel_runs  2 / 5 active
  ✓ dark_launch    not configured
─────────────────────────────────────
  verdict: allow

RPC called: fn_evaluate_pre_run_policy(p_ai_lenser_id, p_workflow_id, p_context)

---

Budget

Usage

lf budget set|status <handle>

lf budget set

lf budget set @my-lenser --ceiling 50.00 --period monthly

Budget ceiling updated for @my-lenser
  ceiling:  $50.00
  period:   monthly
  current spend this period: $12.40

Flags

Flag	Description
--ceiling <usd>	Maximum spend in USD before runs are denied
--period <monthly|daily|rolling-30d>	Budget reset window

lf budget status

lf budget status @my-lenser

Budget status for @my-lenser
  ceiling:    $50.00 (monthly)
  spent:      $12.40
  remaining:  $37.60
  resets at:  2026-06-01T00:00:00Z

RPC called: fn_set_workspace_setting(p_workspace_id, p_key, p_value)

---

Dark Launch

Usage

lf dark-launch enable|disable|status <handle>

lf dark-launch enable

lf dark-launch enable @my-lenser --workflow wf_abc123 --percentage 10

Dark launch enabled for workflow wf_abc123
  percentage:  10%
  routing:     md5(workflow_id) % 100 < 10
  in window:   deterministic (same result on every check)

lf dark-launch disable

lf dark-launch disable @my-lenser --workflow wf_abc123

Dark launch disabled for workflow wf_abc123
  all runs for this workflow will now proceed normally

lf dark-launch status

lf dark-launch status @my-lenser

Dark launch status for @my-lenser
  workflow wf_abc123   10%  ACTIVE
  workflow wf_def456   50%  ACTIVE

Flags

Flag	Description
--workflow <id>	Target workflow ID
--percentage <0-100>	Percentage of runs to allow through

RPC called: fn_set_dark_launch(p_ai_lenser_id, p_workflow_id, p_percentage)

---

Policy Log

Usage

lf policy log|stats <handle>

lf policy log

Stream recent policy evaluation rows (most recent first):

lf policy log @my-lenser --limit 5

Policy evaluations for @my-lenser (last 5)
  2026-05-01 14:22:01  pre_run               allow               –
  2026-05-01 14:21:44  pre_run               deny                kill_switch_active
  2026-05-01 14:18:03  pre_tool_invocation   allow               –
  2026-05-01 14:17:55  pre_run               allow               –
  2026-05-01 14:10:12  pre_run               deny                budget_ceiling_exceeded

Flags for lf policy log

Flag	Description
--limit <n>	Number of rows to return (default 20)
--since <ISO date>	Only show evaluations after this timestamp
--verdict <verdict>	Filter by verdict (allow, deny, pause, require_approval)

lf policy stats

Aggregate verdict counts for a time window:

lf policy stats @my-lenser --period 7d

Policy stats for @my-lenser (last 7 days)
  allow              142
  deny                 8   (budget_ceiling_exceeded: 5, kill_switch_active: 3)
  pause                0
  require_approval     2
  ─────────────────────
  total              152

Flags for lf policy stats

Flag	Description
--period <1d|7d|30d>	Aggregation window (default 7d)

---

Related

• Autonomous Agent OS — governance controls overview
• Policy Engine — RPC spec and verdict definitions
• Run Reports & Incidents — post-run artifact reference
• Using the Kill Switch — step-by-step operator guide
• Dark Launch Rollout — ramp procedure