Fix #2012: gate agent-heartbeat-stall overseer alerts on Tier-1 signal

sandbox/agent-config/rules/overseer.md

@@ -106,6 +106,7 @@ Sonnet calls use `model="sonnet"` (or the configured `overseer_decision_maker_mo
 Not all silence means stalling. Distinguish between:
 
 - **Legitimate long-running work**: Compilation, large test suites, complex refactoring. Look for tool call activity, file changes, and process output even when progress events are sparse.
+- **Reviewer waiting on a producer (BRC)**: In BRC phases (refine, plan, implement), reviewer-only roles are expected to be silent while the producer is in `WORKING`. They have literally nothing to review until the producer emits `PROPOSED`. The Tier-1 monitor already suppresses heartbeat/progress alerts for reviewers in this state (`_is_brc_idle` in `orchestrator/health_monitor.py:177-218`). **Never synthesize a stall alert against a reviewer-only role whose upstream producer is still `WORKING`** — this is expected behavior, not a stall. Refine-phase producer WORKING commonly runs 6–10 minutes before `PROPOSED`, especially when the contract has many unresolved decisions.
 - **Genuinely stuck**: No tool calls, no file changes, no progress events. The agent may be looping, waiting on a resource, or confused by an error.
 - **Needs help**: The agent is active but producing errors or making no meaningful progress. It may need a redirect or additional context.
 
@@ -172,11 +173,11 @@ Emit an `OVERSEER_ALERT` when you observe any of these:
 - **Stuck phase transition**: BRC consensus is `complete` (`consensus.state == "confirmed"`) but the phase has not transitioned within ~60 seconds. Anomaly type: `stuck-phase-transition`, priority: `high`. Include the consensus state and the time since confirmation in `--detail`. **Do NOT emit `stuck-phase-transition` when `consensus.state != "confirmed"`** — that is a different failure mode (producer still working, reviewers still ACKing, etc.) and requires a different anomaly type. An empty contract during the refine phase is never grounds for this alert; see [Phase-relative baselines](#phase-relative-baselines).
 - **Orchestrator silent on consensus**: No `CONSENSUS_*` messages from the orchestrator for several minutes while agents are still active. Anomaly type: `orchestrator-consensus-silent`, priority: `high`.
 - **Repeated 401 from any orchestrator endpoint**: You tried a command and got a 401 (or any auth-rejection). **Stop retrying that command immediately.** Anomaly type: `unauthorized-overseer-action`, priority: `medium`. Include which command you tried and why you thought it was needed.
-- **Heartbeat stall on an active agent**: An agent has missed `max_missed_heartbeats` consecutive heartbeats **and** the health alert from the orchestrator confirms it **and** the phase-specific minimum working window (see [Phase-relative baselines](#phase-relative-baselines)) has elapsed. Anomaly type: `agent-heartbeat-stall`, priority depends on the agent's criticality. A producer that has been `WORKING` for less than the phase floor without a `CONSENSUS_PROPOSE` is **not** stalled. Note: active tool-call activity (as described in [Phase-relative baselines](#phase-relative-baselines)) is relevant counter-evidence even outside the first-proposal context — if the agent is producing tool calls, weigh that against the missed-heartbeat signal before alerting.
+- **Heartbeat stall on an active agent**: Fire **only** when the orchestrator's Tier-1 heartbeat monitor has already tripped — i.e. `alerts_detail` contains an entry with `alert_type` of `heartbeat_timeout` (or similar heartbeat-related type) for the agent in question — **and** the phase-specific minimum working window (see [Phase-relative baselines](#phase-relative-baselines)) has elapsed. If no such alert is present, the system-level threshold has **not** been crossed and you **must not** synthesize an `agent-heartbeat-stall` alert from your own observation of quiet roles. An alert whose own body says "system threshold not yet crossed" or "no action required" is a false positive by definition — do not emit it. Note: active tool-call activity (as described in [Phase-relative baselines](#phase-relative-baselines)) is relevant counter-evidence even when a heartbeat alert is present — if the agent is producing tool calls, weigh that against the heartbeat signal before alerting. Anomaly type: `agent-heartbeat-stall`, priority depends on the agent's criticality.
 - **Persistent agent loop**: Haiku classifier returns `loop` with confidence > 0.8 across two consecutive cycles. Anomaly type: `agent-loop`, priority: `medium`.
 - **Same anomaly seen across N cycles without resolution**: If you've already alerted on the same anomaly and the situation hasn't changed for `overseer_max_cycles_before_re_alert` (default: 3) cycles, re-alert with priority bumped one level.
 
-When in doubt: alert — a spurious alert is cheaper than a silent stuck pipeline. **But every alert must cite specific observed evidence** (log line, message ID, progress-event reference, `consensus.state` value, elapsed-time figure) that contradicts the "working normally" null hypothesis. "The contract is empty" or "no CONSENSUS_PROPOSE yet" are not evidence on their own; pair them with phase-relative context (see [Phase-relative baselines](#phase-relative-baselines)) before escalating.
+When in doubt on a qualitative signal (error content, agent misalignment, mediation need): alert. A spurious alert is cheaper than a silent stuck pipeline. **But every alert must cite specific observed evidence** (log line, message ID, progress-event reference, `consensus.state` value, elapsed-time figure) that contradicts the "working normally" null hypothesis. "The contract is empty" or "no CONSENSUS_PROPOSE yet" are not evidence on their own; pair them with phase-relative context (see [Phase-relative baselines](#phase-relative-baselines)) before escalating. **Exception**: do not apply this bias to anomalies that have deterministic JSON triggers listed above — `agent-heartbeat-stall`, `stuck-phase-transition`, `orchestrator-consensus-silent`, `agent-loop`. Those require the trigger condition (specific `alerts_detail`/consensus/classifier values) to actually hold. Do not fire them on a hunch.
 
 ## OVERSEER_ALERT body format