Known Issues with Claude Flow Benchmark System

Issue 1: Non-Interactive Mode Not Working in Claude Flow

Problem

The --non-interactive flag is not being properly handled by claude-flow commands, particularly:

• hive-mind spawn --non-interactive still prompts for interactive input
• swarm --non-interactive may also have similar issues

Current Status

The benchmark system correctly passes the --non-interactive flag in the proper command format:

./claude-flow hive-mind spawn "Task description" --count 2 --coordination majority --non-interactive

However, the claude-flow implementation needs to be updated to:

1. Check for the --non-interactive flag before calling wizard functions
2. Use default values or passed parameters instead of prompting

Workaround

Currently, SPARC commands work correctly without requiring interactive input:

swarm-benchmark real sparc tdd "Create a function"

Fix Required in Claude Flow

The following files need updates:

• /src/cli/simple-commands/hive-mind.js - Check for non-interactive flag in spawn function
• /src/cli/simple-commands/swarm.js - Similar check needed

Example fix needed:

async function spawnSwarm(args, flags) {
  // Check for non-interactive mode FIRST
  if (flags['non-interactive'] || flags.nonInteractive) {
    // Use defaults or passed parameters without prompting
    const objective = args.join(' ').trim();
    if (!objective) {
      console.error('Objective required in non-interactive mode');
      return;
    }
    // Proceed with defaults...
  } else if (flags.wizard || subArgs.length === 0) {
    await spawnSwarmWizard();
  }
  // ...
}

Issue 2: Command Timeout Handling

Problem

Long-running benchmarks need proper timeout configuration.

Solution

Timeouts have been increased to:

• Swarm: Uses configured timeout (default 6 hours)
• Hive-mind: 6 hours
• SPARC: 2 hours

These can be adjusted in the benchmark configuration.

Issue 3: Claude CLI Dependency

Problem

Swarm and hive-mind commands require Claude CLI even with --executor flag.

Current Behavior

Commands will execute but may fail if Claude CLI is not properly configured.

Workaround

Use SPARC commands for benchmarking without Claude CLI dependency.