/**
* Auto-Verify Prompt
*
* Instructions for automatic verification using MCP Sampling
*/
export function generateAutoVerifyPrompt(args: Record<string, string>): string {
const target = args.target || '.';
const requirements = args.requirements || 'General code quality, security, and correctness verification';
const maxRounds = args.maxRounds || '10';
return `# Elenchus Auto-Verification
You are about to run an **AUTOMATIC** verification loop using MCP Sampling.
## What This Does
The server will autonomously orchestrate a Verifier↔Critic loop:
1. **Verifier Round**: Server generates prompt, requests LLM completion, extracts issues
2. **Critic Round**: Server generates review prompt, requests LLM completion, validates issues
3. **Repeat**: Until convergence or max rounds reached
**You don't need to manually run rounds** - the server handles everything via Sampling.
## How to Start
Call the \`elenchus_auto_verify\` tool:
\`\`\`json
{
"target": "${target}",
"requirements": "${requirements}",
"workingDir": "${process.cwd()}",
"config": {
"maxRounds": ${maxRounds},
"stopOnCritical": false,
"autoConsolidate": true
}
}
\`\`\`
## Configuration Options
| Option | Default | Description |
|--------|---------|-------------|
| \`maxRounds\` | 10 | Maximum verification rounds |
| \`maxTokens\` | 4000 | Tokens per LLM request |
| \`stopOnCritical\` | false | Stop on first CRITICAL issue |
| \`minRounds\` | 2 | Minimum rounds before convergence |
| \`autoConsolidate\` | true | Generate fix plan at end |
## What You Get Back
\`\`\`typescript
{
sessionId: string;
status: 'converged' | 'stopped' | 'error';
totalRounds: number;
duration: number; // milliseconds
issues: {
total: number;
critical: number;
high: number;
resolved: number;
};
convergenceReason?: string;
consolidatedPlan?: {
mustFix: Issue[];
shouldFix: Issue[];
couldFix: Issue[];
wontFix: Issue[];
totalEffort: string;
};
}
\`\`\`
## Requirements
**Client must support MCP Sampling capability** for this to work.
If your client doesn't support Sampling, use the manual \`/verify\` command instead.
## Target Information
- **Target**: ${target}
- **Requirements**: ${requirements}
- **Max Rounds**: ${maxRounds}
---
**Start the auto-verification now by calling \`elenchus_auto_verify\`.**
`;
}
