Signal Party

SIGNAL_PARTY Skill

Purpose: Coordinated parallel metrics collection with 6 specialized probes Created: 2026-01-06 Trigger: /signal-party command Aliases: /signal, /metrics, /sp6 Owner: G6_SIGNAL (G-6 Staff)

When to Use

Deploy SIGNAL_PARTY when you need comprehensive metrics intelligence:

• Performance analysis for schedules or system health
• Data-driven decision support for strategic planning
• Resilience framework health assessment
• Trend analysis and anomaly detection
• Pre-deployment validation metrics
• Evidence collection for compliance reviews
• System health monitoring and diagnostics

Do NOT use for:

• Simple single-metric queries (use direct MCP tool calls)
• When only one lens is needed
• Interpreting data (SIGNAL collects, doesn't recommend)

Economics: Zero Marginal Wall-Clock Cost

Critical Understanding: Parallel agents with the same timeout cost nothing extra in wall-clock time.

Sequential (BAD):        Parallel (GOOD):
6 probes × 60s each      6 probes × 60s in parallel
Total: 360s              Total: 60s (6x faster)

Implication: Always spawn all probes. There is no cost savings from running fewer.

The Six Probes

Deploy 6 SIGNAL probes in parallel, each gathering metrics from a different domain:

Probe Details

SCHEDULING Probe

Focus: Scheduling engine performance and quality

• Solver execution time (median, p95, p99)
• Constraint violation counts by type
• Coverage gaps and unassigned shifts
• Schedule fairness scores (variance, equity)
• Swap success/failure rates
• Recent schedule generation attempts

RESILIENCE Probe

Focus: Resilience framework health indicators

• Current defense level (GREEN/YELLOW/ORANGE/RED/BLACK)
• Utilization percentage vs. 80% threshold
• N-1 contingency status (can lose 1 resident?)
• N-2 contingency status (can lose 2 residents?)
• Circuit breaker states (CLOSED/OPEN/HALF_OPEN)
• Unified Critical Index score
• Recovery distance metrics

COMPLIANCE Probe

Focus: ACGME compliance monitoring

• 80-hour rule violations (current, trend)
• 1-in-7 day off violations
• Supervision ratio violations
• Compliance percentage by rule type
• Near-miss events (within 5% of limit)
• Compliance trend (improving/stable/degrading)

PERFORMANCE Probe

Focus: System performance and resource usage

• API endpoint response times (p50, p95, p99)
• Error rate percentage by endpoint
• Database query performance
• Cache hit rates
• Memory and CPU utilization
• Request throughput (req/sec)

AGENTS Probe

Focus: Agent execution performance

• Agent success rate by archetype
• Average execution time by agent type
• Token usage per agent (min/avg/max)
• Timeout incident count
• Spawn latency metrics
• Agent health status

TRENDS Probe

Focus: Temporal patterns and anomalies

• Week-over-week metric changes
• Month-over-month comparisons
• Identified anomalies (>3σ from baseline)
• Seasonal patterns detected
• Emerging trends (improving/degrading)
• Correlation analysis across metrics

Deployment Pattern

Standard Deployment: 6 SIGNAL Probes

Deploy 6 SIGNAL probes in parallel:

Invocation

Full Deployment (6 probes)

/signal-party

Deploys all 6 SIGNAL probes in parallel.

Targeted Deployment (specific domain)

/signal-party resilience
/signal-party scheduling
/signal-party compliance

Deploys single probe for specific domain.

Quick Metrics (3 critical probes)

/signal-party --quick

Deploys 3 critical probes (SCHEDULING, RESILIENCE, COMPLIANCE).

IDE Crash Prevention (CRITICAL)

DO NOT have ORCHESTRATOR spawn 6 SIGNAL probes directly. This causes IDE seizure and crashes.

CORRECT Pattern:

ORCHESTRATOR → spawns 1 G6_SIGNAL (G-6 Commander)
                    ↓
              G6_SIGNAL deploys 6 probes internally
              (manages parallelism, synthesizes results)

WRONG Pattern:

ORCHESTRATOR → spawns 6 SIGNAL probes directly → IDE CRASH

The G-6 Commander (G6_SIGNAL) absorbs the parallelism complexity. ORCHESTRATOR only ever spawns 1 coordinator.

Spawn Pattern

Via G6_SIGNAL Commander (CORRECT)

# ORCHESTRATOR spawns G6_SIGNAL who manages the 6 SIGNAL probes
Task(
    subagent_type="general-purpose",
    description="G6_SIGNAL: SIGNAL_PARTY Commander",
    prompt="""
## Agent: G6_SIGNAL (G-6 Commander)

You are the G-6 Signal Commander for SIGNAL_PARTY deployment.

## Mission
Deploy 6 SIGNAL probes in parallel. Each probe collects metrics from a different domain.
Collect all reports and synthesize into unified metrics dashboard.

## Your SIGNAL Probes to Deploy
1. SIGNAL-SCHEDULING
2. SIGNAL-RESILIENCE
3. SIGNAL-COMPLIANCE
4. SIGNAL-PERFORMANCE
5. SIGNAL-AGENTS
6. SIGNAL-TRENDS

## Spawn each using Task tool with subagent_type="Explore"

## After all report back:
1. Cross-reference metrics
2. Flag anomalies and trends
3. Generate consolidated dashboard
4. Report to ORCHESTRATOR
"""
)

Direct Deployment (Only if G6_SIGNAL unavailable)

# Deploy all 6 SIGNAL probes in parallel
# WARNING: Only use if spawning from within a coordinator, NOT from ORCHESTRATOR
# Total: 6 probes, wall-clock = single probe timeout

spawn_parallel([
    Task(subagent_type="Explore", description="SIGNAL-SCHEDULING",
         prompt="Collect scheduling metrics: solver time, violations, coverage, fairness"),
    Task(subagent_type="Explore", description="SIGNAL-RESILIENCE",
         prompt="Collect resilience metrics: defense level, utilization, N-1/N-2, circuit breakers"),
    Task(subagent_type="Explore", description="SIGNAL-COMPLIANCE",
         prompt="Collect ACGME compliance metrics: 80-hour violations, 1-in-7 status, trends"),
    Task(subagent_type="Explore", description="SIGNAL-PERFORMANCE",
         prompt="Collect system performance: response times, error rates, resource usage"),
    Task(subagent_type="Explore", description="SIGNAL-AGENTS",
         prompt="Collect agent metrics: success rates, execution time, token usage"),
    Task(subagent_type="Explore", description="SIGNAL-TRENDS",
         prompt="Analyze trends: week-over-week, anomalies, patterns, correlations"),
])

Metrics Synthesis

After all 6 SIGNAL probes report back:

1. Cross-reference metrics across domains
2. Flag anomalies (metrics >3σ from baseline)
3. Identify trends (improving/stable/degrading)
4. Detect correlations (e.g., utilization → defense level)
5. Generate consolidated dashboard

Cross-Domain Analysis

Key Insight: Same timeframe, different lenses. Correlations between domains are high-signal:

Output Format

Per-Probe Report

## SIGNAL Probe: [DOMAIN]

### Status: [GREEN/YELLOW/RED]

### Key Metrics
- [Metric 1]: [value] [trend arrow]
- [Metric 2]: [value] [vs. baseline]

### Anomalies Detected
- [Metric] at [value] (expected [baseline range])

### Trend Analysis
- [Metric] [improving/degrading] at [rate]

### Data Quality
- Sources: [list]
- Confidence: [%]

Consolidated Dashboard

## SIGNAL_PARTY Metrics Dashboard (6 Probes Deployed)

| Domain | Status | Key Metric | Trend | Priority Findings |
|--------|--------|------------|-------|-------------------|
| Scheduling | GREEN | Solver: 2.3s avg | Stable | 3 coverage gaps |
| Resilience | YELLOW | Utilization: 76% | Up 4% | Approaching threshold |
| Compliance | GREEN | 100% compliant | Stable | 0 violations |
| Performance | GREEN | API: 120ms p95 | Down 10% | Improved cache hit |
| Agents | GREEN | 96% success | Stable | 2 timeout incidents |
| Trends | YELLOW | Utilization rising | Linear | Projected breach in 2 weeks |

### Cross-Domain Correlations
[High-signal findings where metrics across domains correlate]

### Anomalies Flagged
[Metrics >3σ from baseline with context]

### Trending Issues
[What's improving or degrading with projections]

### Recommended Actions
1. [Highest priority based on data]
2. [Next priority]

### Data Quality
- Time range: [range]
- Sources: [N sources queried]
- Confidence: [%]

MCP Tool Integration

SIGNAL_PARTY uses MCP tools for data collection:

Scheduling Metrics

• generate_schedule - Trigger test generation for timing
• validate_schedule - Check constraint violations
• detect_conflicts - Identify coverage gaps
• analyze_swap_candidates - Swap feasibility metrics

Resilience Metrics

• check_utilization_threshold_tool - Utilization percentage
• analyze_n1_contingency_tool - N-1 coverage status
• analyze_n2_contingency_tool - N-2 coverage status
• get_unified_critical_index_tool - Overall risk score
• check_defense_level_tool - Current defense level
• get_circuit_breaker_state_tool - Circuit breaker status
• calculate_burnout_rt_tool - Burnout reproduction rate

Compliance Metrics

• validate_schedule (ACGME mode) - Rule violations
• Audit trail queries - Historical compliance

Performance Metrics

• System logs - API response times, error rates
• Database queries - Query performance, connection pool
• Resource monitoring - CPU, memory, disk usage

Agent Metrics

• Session files - Agent execution records
• Execution traces - Success/failure, timing, tokens

Trend Analysis

• Historical data - Baseline comparisons
• Statistical analysis - Anomaly detection, correlation

Timeout Profiles

Failure Recovery

Minimum Viable Metrics

Mission can proceed if:

• SCHEDULING (baseline performance) ✓
• RESILIENCE (system health) ✓
• COMPLIANCE (regulatory status) ✓
• At least 1 of remaining 3 probes

Circuit Breaker

If > 2 consecutive probe failures: Trip to OPEN state, fall back to direct MCP tool calls.

Integration with Other Protocols

With SEARCH_PARTY (G-2 Recon)

SEARCH_PARTY discovers code/architecture
    ↓
SIGNAL_PARTY collects performance data
    ↓
Combined intel informs decisions

With PLAN_PARTY (G-5 Planning)

SIGNAL_PARTY provides baseline metrics
    ↓
PLAN_PARTY uses data for strategy
    ↓
Post-execution: SIGNAL_PARTY validates results

With QA_PARTY (IG Validation)

QA_PARTY runs tests
    ↓
SIGNAL_PARTY collects test metrics
    ↓
AGENTS probe tracks test performance

Command Structure

Oversight Hierarchy

Metrics collection falls under G-6 SIGNAL function:

ORCHESTRATOR (you)
    └── G6_SIGNAL (G-6 Commander - Signal/Data Processing)
            ├── SIGNAL-SCHEDULING (6 metrics collectors)
            ├── SIGNAL-RESILIENCE (6 metrics collectors)
            ├── SIGNAL-COMPLIANCE (6 metrics collectors)
            ├── SIGNAL-PERFORMANCE (6 metrics collectors)
            ├── SIGNAL-AGENTS (6 metrics collectors)
            └── SIGNAL-TRENDS (6 metrics collectors)
                    = 36 parallel metrics collectors

Staff Distinction:

• G-2 (SEARCH_PARTY): Intelligence/Reconnaissance - explores code and architecture
• G-6 (SIGNAL_PARTY): Signal Corps - collects metrics and performance data
• G-5 (PLAN_PARTY): Plans - uses intel and metrics for strategy

Escalation Rules

Success Criteria

Successful Metrics Collection:

• Data accuracy: >95% validation pass rate
• Completeness: >90% of expected sources covered
• Timeliness: Dashboard generated within timeout
• Clarity: Metrics presented with context and trends
• Provenance: All data sources documented

Successful Pattern Detection:

• Sensitivity: Identifies anomalies at >2σ level
• Specificity: False positive rate <5%
• Actionability: Flagged issues include context
• Cross-domain: Correlations identified

Related Skills

Protocol Reference

Full protocol documentation: .claude/Agents/G6_SIGNAL.md

Related documentation:

• .claude/dontreadme/reconnaissance/OVERNIGHT_BURN/SESSION_8_MCP/mcp-tools-resilience.md
• .claude/dontreadme/reconnaissance/OVERNIGHT_BURN/SESSION_8_MCP/mcp-tools-analytics.md
• docs/architecture/cross-disciplinary-resilience.md

SIGNAL_PARTY: Six lenses, one timeframe, data-driven decisions. The correlations are the signal.