You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

step-04-evaluate-and-score.md 6.9KB


name: ‘step-04-evaluate-and-score’ description: ‘Orchestrate adaptive NFR evidence domain audits (agent-team, subagent, or sequential)’

nextStepFile: ‘{skill-root}/steps-c/step-04e-aggregate-nfr.md’

Step 4: Orchestrate Adaptive NFR Evidence Audit

STEP GOAL

Select execution mode deterministically, then audit NFR evidence domains using agent-team, subagent, or sequential execution while preserving output contracts.

MANDATORY EXECUTION RULES

  • 📖 Read the entire step file before acting
  • ✅ Speak in {communication_language}
  • ✅ Resolve execution mode from config (tea_execution_mode, tea_capability_probe)
  • ✅ Apply fallback rules deterministically when requested mode is unsupported
  • ✅ Wait for required worker steps to complete
  • ❌ Do NOT skip capability checks when probing is enabled

EXECUTION PROTOCOLS:

  • 🎯 Follow the MANDATORY SEQUENCE exactly
  • 💾 Wait for subagent outputs
  • 📖 Load the next step only when instructed

MANDATORY SEQUENCE

1. Prepare Execution Context

Generate unique timestamp:

const timestamp = new Date().toISOString().replace(/[:.]/g, '-');

Prepare context:

const parseBooleanFlag = (value, defaultValue = true) => {
  if (typeof value === 'string') {
    const normalized = value.trim().toLowerCase();
    if (['false', '0', 'off', 'no'].includes(normalized)) return false;
    if (['true', '1', 'on', 'yes'].includes(normalized)) return true;
  }
  if (value === undefined || value === null) return defaultValue;
  return Boolean(value);
};

const subagentContext = {
  system_context: /* from Step 1 */,
  nfr_thresholds: /* from Step 2 */,
  evidence_gathered: /* from Step 3 */,
  config: {
    execution_mode: config.tea_execution_mode || 'auto',  // "auto" | "subagent" | "agent-team" | "sequential"
    capability_probe: parseBooleanFlag(config.tea_capability_probe, true),  // supports booleans and "false"/"true" strings
  },
  timestamp: timestamp
};

2. Resolve Execution Mode with Capability Probe

const normalizeUserExecutionMode = (mode) => {
  if (typeof mode !== 'string') return null;
  const normalized = mode.trim().toLowerCase().replace(/[-_]/g, ' ').replace(/\s+/g, ' ');

  if (normalized === 'auto') return 'auto';
  if (normalized === 'sequential') return 'sequential';
  if (normalized === 'subagent' || normalized === 'sub agent' || normalized === 'subagents' || normalized === 'sub agents') {
    return 'subagent';
  }
  if (normalized === 'agent team' || normalized === 'agent teams' || normalized === 'agentteam') {
    return 'agent-team';
  }

  return null;
};

const normalizeConfigExecutionMode = (mode) => {
  if (mode === 'subagent') return 'subagent';
  if (mode === 'auto' || mode === 'sequential' || mode === 'subagent' || mode === 'agent-team') {
    return mode;
  }
  return null;
};

// Explicit user instruction in the active run takes priority over config.
const explicitModeFromUser = normalizeUserExecutionMode(runtime.getExplicitExecutionModeHint?.() || null);

const requestedMode = explicitModeFromUser || normalizeConfigExecutionMode(subagentContext.config.execution_mode) || 'auto';
const probeEnabled = subagentContext.config.capability_probe;

const supports = {
  subagent: false,
  agentTeam: false,
};

if (probeEnabled) {
  supports.subagent = runtime.canLaunchSubagents?.() === true;
  supports.agentTeam = runtime.canLaunchAgentTeams?.() === true;
}

let resolvedMode = requestedMode;

if (requestedMode === 'auto') {
  if (supports.agentTeam) resolvedMode = 'agent-team';
  else if (supports.subagent) resolvedMode = 'subagent';
  else resolvedMode = 'sequential';
} else if (probeEnabled && requestedMode === 'agent-team' && !supports.agentTeam) {
  resolvedMode = supports.subagent ? 'subagent' : 'sequential';
} else if (probeEnabled && requestedMode === 'subagent' && !supports.subagent) {
  resolvedMode = 'sequential';
}

subagentContext.execution = {
  requestedMode,
  resolvedMode,
  probeEnabled,
  supports,
};

Resolution precedence:

  1. Explicit user request in this run (agent team => agent-team; subagent => subagent; sequential; auto)
  2. tea_execution_mode from config
  3. Runtime capability fallback (when probing enabled)

If probing is disabled, honor the requested mode strictly. If that mode cannot be executed at runtime, fail with explicit error instead of silent fallback.


3. Dispatch 4 NFR Workers

Subagent A: Security Evidence Audit

  • File: ./step-04a-subagent-security.md
  • Output: /tmp/tea-nfr-security-${timestamp}.json
  • Execution:
    • agent-team or subagent: launch non-blocking
    • sequential: run blocking and wait
  • Status: Running… ⟳

Subagent B: Performance Evidence Audit

  • File: ./step-04b-subagent-performance.md
  • Output: /tmp/tea-nfr-performance-${timestamp}.json
  • Status: Running… ⟳

Subagent C: Reliability Evidence Audit

  • File: ./step-04c-subagent-reliability.md
  • Output: /tmp/tea-nfr-reliability-${timestamp}.json
  • Status: Running… ⟳

Subagent D: Scalability Evidence Audit

  • File: ./step-04d-subagent-scalability.md
  • Output: /tmp/tea-nfr-scalability-${timestamp}.json
  • Status: Running… ⟳

In agent-team and subagent modes, runtime decides worker scheduling and concurrency.


4. Wait for Expected Worker Completion

If resolvedMode is agent-team or subagent:

⏳ Waiting for 4 NFR subagents to complete...
  ├── Subagent A (Security): Running... ⟳
  ├── Subagent B (Performance): Running... ⟳
  ├── Subagent C (Reliability): Running... ⟳
  └── Subagent D (Scalability): Running... ⟳

[... time passes ...]

✅ All 4 NFR subagents completed!

If resolvedMode is sequential:

✅ Sequential mode: each worker already completed during dispatch.

5. Verify All Outputs Exist

const outputs = ['security', 'performance', 'reliability', 'scalability'].map((domain) => `/tmp/tea-nfr-${domain}-${timestamp}.json`);

outputs.forEach((output) => {
  if (!fs.existsSync(output)) {
    throw new Error(`Subagent output missing: ${output}`);
  }
});

6. Execution Report

🚀 Performance Report:
- Execution Mode: {resolvedMode}
- Total Elapsed: ~mode-dependent
- Parallel Gain: ~67% faster when mode is subagent/agent-team

7. Proceed to Aggregation

Load next step: {nextStepFile}

The aggregation step will:

  • Read all 4 NFR domain outputs
  • Calculate overall risk level
  • Aggregate compliance status
  • Identify cross-domain risks
  • Generate executive summary

EXIT CONDITION

Proceed when all 4 required worker steps completed and outputs exist.


🚨 SYSTEM SUCCESS METRICS

✅ SUCCESS:

  • All required worker steps completed
  • Fallback behavior respected configuration and capability probe rules

❌ FAILURE:

  • One or more subagents failed
  • Unsupported requested mode with probing disabled