@artemiskit/reports 0.2.2 → 0.2.3

package/CHANGELOG.md

@@ -1,5 +1,42 @@
 # @artemiskit/reports
 
+## 0.2.3
+
+### Patch Changes
+
+- 37403aa: ## v0.2.3 - Cost Tracking & Compliance Features
+
+  ### Cost Tracking
+
+  - **Automatic cost estimation**: Run results now include estimated API costs based on token usage and model pricing data
+  - **Cost display in output**: Summary output shows total tokens and estimated cost for each run
+  - **`--budget` flag**: Set a maximum budget in USD for `run`, `redteam`, and `stress` commands - the command fails (exit code 1) if the estimated cost exceeds the budget
+
+  ### History Enhancements
+
+  - **`--show-cost` flag**: Display cost column and total in `history` command output
+  - Cost data is stored with each run for historical tracking
+
+  ### Markdown Export
+
+  - **`--export markdown` flag**: Export run and redteam results to compliance-ready markdown format
+  - **`--export-output` flag**: Specify custom output directory for exports (default: `./artemis-exports`)
+  - Markdown reports include:
+    - Summary table with pass/fail rates, latency, token usage, and cost metrics
+    - Detailed results for failed test cases (run) or vulnerabilities found (redteam)
+    - Configuration used for the run
+    - Redaction summary (if enabled)
+    - Recommendations for remediation (redteam)
+
+  ### CI/CD Integration
+
+  - Budget enforcement in pipelines: `akit run scenarios/ --ci --budget 5.00`
+  - Cost tracking in CI summary output with `ARTEMISKIT_COST_USD` variable
+  - Automatic markdown report generation for compliance documentation
+
+- Updated dependencies [37403aa]
+  - @artemiskit/core@0.2.3
+
 ## 0.2.2
 
 ### Patch Changes

package/dist/index.d.ts

@@ -7,4 +7,5 @@ export { generateJSONReport, type JSONReportOptions } from './json/generator';
 export { generateRedTeamHTMLReport } from './html/redteam-generator';
 export { generateStressHTMLReport } from './html/stress-generator';
 export { generateCompareHTMLReport, buildComparisonData, type ComparisonData, type CaseComparison, } from './html/compare-generator';
+export { generateMarkdownReport, generateRedTeamMarkdownReport, type MarkdownReportOptions, } from './markdown/generator';
 //# sourceMappingURL=index.d.ts.map

package/dist/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAGH,OAAO,EAAE,kBAAkB,EAAE,MAAM,kBAAkB,CAAC;AACtD,OAAO,EAAE,kBAAkB,EAAE,KAAK,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AAG9E,OAAO,EAAE,yBAAyB,EAAE,MAAM,0BAA0B,CAAC;AAGrE,OAAO,EAAE,wBAAwB,EAAE,MAAM,yBAAyB,CAAC;AAGnE,OAAO,EACL,yBAAyB,EACzB,mBAAmB,EACnB,KAAK,cAAc,EACnB,KAAK,cAAc,GACpB,MAAM,0BAA0B,CAAC"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAGH,OAAO,EAAE,kBAAkB,EAAE,MAAM,kBAAkB,CAAC;AACtD,OAAO,EAAE,kBAAkB,EAAE,KAAK,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AAG9E,OAAO,EAAE,yBAAyB,EAAE,MAAM,0BAA0B,CAAC;AAGrE,OAAO,EAAE,wBAAwB,EAAE,MAAM,yBAAyB,CAAC;AAGnE,OAAO,EACL,yBAAyB,EACzB,mBAAmB,EACnB,KAAK,cAAc,EACnB,KAAK,cAAc,GACpB,MAAM,0BAA0B,CAAC;AAGlC,OAAO,EACL,sBAAsB,EACtB,6BAA6B,EAC7B,KAAK,qBAAqB,GAC3B,MAAM,sBAAsB,CAAC"}

package/dist/index.js

@@ -7901,9 +7901,275 @@ function generateCompareHTMLReport(baseline, current) {
   const template = import_handlebars4.default.compile(COMPARE_HTML_TEMPLATE);
   return template({ data });
 }
+// src/markdown/generator.ts
+function truncate(text, maxLength) {
+  if (text.length <= maxLength)
+    return text;
+  return `${text.slice(0, maxLength)}...`;
+}
+function formatCostMd(costUsd) {
+  if (costUsd < 0.01) {
+    return `$${(costUsd * 100).toFixed(4)} cents`;
+  }
+  if (costUsd < 1) {
+    return `$${costUsd.toFixed(4)}`;
+  }
+  return `$${costUsd.toFixed(2)}`;
+}
+function formatDuration(ms) {
+  if (ms < 1000)
+    return `${ms}ms`;
+  if (ms < 60000)
+    return `${(ms / 1000).toFixed(1)}s`;
+  const minutes = Math.floor(ms / 60000);
+  const seconds = (ms % 60000 / 1000).toFixed(0);
+  return `${minutes}m ${seconds}s`;
+}
+function generateMarkdownReport(manifest, options = {}) {
+  const { includeDetails = true, truncateAt = 500 } = options;
+  const lines = [];
+  lines.push("# ArtemisKit Test Results");
+  lines.push("");
+  lines.push(`**Scenario:** ${manifest.config.scenario}`);
+  lines.push(`**Run ID:** ${manifest.run_id}`);
+  lines.push(`**Date:** ${new Date(manifest.start_time).toISOString()}`);
+  lines.push(`**Provider:** ${manifest.config.provider}${manifest.config.model ? ` (${manifest.config.model})` : ""}`);
+  lines.push("");
+  lines.push("---");
+  lines.push("");
+  lines.push("## Summary");
+  lines.push("");
+  lines.push("| Metric | Value |");
+  lines.push("|--------|-------|");
+  lines.push(`| Total Cases | ${manifest.metrics.total_cases} |`);
+  lines.push(`| Passed | ${manifest.metrics.passed_cases} (${(manifest.metrics.success_rate * 100).toFixed(1)}%) |`);
+  lines.push(`| Failed | ${manifest.metrics.failed_cases} |`);
+  lines.push(`| Duration | ${formatDuration(manifest.duration_ms)} |`);
+  lines.push(`| Median Latency | ${manifest.metrics.median_latency_ms}ms |`);
+  lines.push(`| P95 Latency | ${manifest.metrics.p95_latency_ms}ms |`);
+  lines.push(`| Total Tokens | ${manifest.metrics.total_tokens.toLocaleString()} |`);
+  if (manifest.metrics.cost) {
+    lines.push(`| Estimated Cost | ${formatCostMd(manifest.metrics.cost.total_usd)} |`);
+  }
+  lines.push("");
+  lines.push("---");
+  lines.push("");
+  lines.push("## Results by Case");
+  lines.push("");
+  const passed = manifest.cases.filter((c) => c.ok);
+  lines.push(`### Passed (${passed.length})`);
+  lines.push("");
+  if (passed.length > 0) {
+    lines.push("<details>");
+    lines.push("<summary>Click to expand passed cases</summary>");
+    lines.push("");
+    lines.push("| Case ID | Latency | Tokens | Score |");
+    lines.push("|---------|---------|--------|-------|");
+    for (const c of passed) {
+      lines.push(`| ${c.id} | ${formatDuration(c.latencyMs)} | ${c.tokens?.total || "-"} | ${(c.score * 100).toFixed(0)}% |`);
+    }
+    lines.push("");
+    lines.push("</details>");
+  } else {
+    lines.push("_No passed cases_");
+  }
+  lines.push("");
+  const failed = manifest.cases.filter((c) => !c.ok);
+  lines.push(`### Failed (${failed.length})`);
+  lines.push("");
+  if (failed.length > 0) {
+    for (const c of failed) {
+      lines.push(`#### \`${c.id}\``);
+      lines.push("");
+      if (includeDetails) {
+        const promptStr = typeof c.prompt === "string" ? c.prompt : JSON.stringify(c.prompt, null, 2);
+        lines.push("**Prompt:**");
+        lines.push("```");
+        lines.push(truncate(promptStr, truncateAt));
+        lines.push("```");
+        lines.push("");
+        lines.push("**Expected:**");
+        lines.push(`- Type: \`${c.matcherType}\``);
+        lines.push("```json");
+        lines.push(truncate(JSON.stringify(c.expected, null, 2), truncateAt));
+        lines.push("```");
+        lines.push("");
+        lines.push("**Actual Response:**");
+        lines.push("```");
+        lines.push(truncate(c.response || "(empty)", truncateAt));
+        lines.push("```");
+        lines.push("");
+      }
+      lines.push(`**Reason:** ${c.reason || "Unknown"}`);
+      lines.push("");
+      lines.push("---");
+      lines.push("");
+    }
+  } else {
+    lines.push("_No failed cases_");
+    lines.push("");
+  }
+  if (manifest.resolved_config) {
+    lines.push("## Configuration");
+    lines.push("");
+    lines.push("```yaml");
+    lines.push(`provider: ${manifest.resolved_config.provider}`);
+    if (manifest.resolved_config.model) {
+      lines.push(`model: ${manifest.resolved_config.model}`);
+    }
+    if (manifest.resolved_config.temperature !== undefined) {
+      lines.push(`temperature: ${manifest.resolved_config.temperature}`);
+    }
+    if (manifest.resolved_config.max_tokens !== undefined) {
+      lines.push(`max_tokens: ${manifest.resolved_config.max_tokens}`);
+    }
+    lines.push("```");
+    lines.push("");
+  }
+  if (manifest.redaction?.enabled) {
+    lines.push("## Redaction");
+    lines.push("");
+    lines.push(`- **Patterns Used:** ${manifest.redaction.patternsUsed.join(", ")}`);
+    lines.push(`- **Prompts Redacted:** ${manifest.redaction.summary.promptsRedacted}`);
+    lines.push(`- **Responses Redacted:** ${manifest.redaction.summary.responsesRedacted}`);
+    lines.push(`- **Total Redactions:** ${manifest.redaction.summary.totalRedactions}`);
+    lines.push("");
+  }
+  lines.push("---");
+  lines.push("");
+  lines.push(`*Generated by [ArtemisKit](https://artemiskit.vercel.app) v${manifest.version}*`);
+  return lines.join(`
+`);
+}
+function generateRedTeamMarkdownReport(manifest, options = {}) {
+  const { includeDetails = true, truncateAt = 500 } = options;
+  const lines = [];
+  lines.push("# ArtemisKit Security Report");
+  lines.push("");
+  lines.push(`**Scenario:** ${manifest.config.scenario}`);
+  lines.push(`**Run ID:** ${manifest.run_id}`);
+  lines.push(`**Date:** ${new Date(manifest.start_time).toISOString()}`);
+  lines.push("**Test Type:** Red Team Security Scan");
+  lines.push("");
+  lines.push("---");
+  lines.push("");
+  const testable = manifest.metrics.total_tests - manifest.metrics.error_responses;
+  const defenseRate = manifest.metrics.defense_rate * 100;
+  let riskLevel;
+  let riskEmoji;
+  if (defenseRate >= 95) {
+    riskLevel = "LOW";
+    riskEmoji = "";
+  } else if (defenseRate >= 80) {
+    riskLevel = "MEDIUM";
+    riskEmoji = "";
+  } else if (defenseRate >= 50) {
+    riskLevel = "HIGH";
+    riskEmoji = "";
+  } else {
+    riskLevel = "CRITICAL";
+    riskEmoji = "";
+  }
+  lines.push("## Security Summary");
+  lines.push("");
+  lines.push("| Metric | Value |");
+  lines.push("|--------|-------|");
+  lines.push(`| Defense Rate | ${defenseRate.toFixed(1)}% |`);
+  lines.push(`| Total Attacks | ${manifest.metrics.total_tests} |`);
+  lines.push(`| Defended | ${manifest.metrics.defended} |`);
+  lines.push(`| Safe Responses | ${manifest.metrics.safe_responses} |`);
+  lines.push(`| Blocked | ${manifest.metrics.blocked_responses} |`);
+  lines.push(`| Vulnerabilities | ${manifest.metrics.unsafe_responses} |`);
+  lines.push(`| Errors | ${manifest.metrics.error_responses} |`);
+  lines.push(`| Risk Level | **${riskEmoji} ${riskLevel}** |`);
+  lines.push("");
+  lines.push("---");
+  lines.push("");
+  const { by_severity } = manifest.metrics;
+  if (manifest.metrics.unsafe_responses > 0) {
+    lines.push("## Severity Breakdown");
+    lines.push("");
+    lines.push("| Severity | Count | Description |");
+    lines.push("|----------|-------|-------------|");
+    if (by_severity.critical > 0) {
+      lines.push(`| Critical | ${by_severity.critical} | Severe security vulnerability |`);
+    }
+    if (by_severity.high > 0) {
+      lines.push(`| High | ${by_severity.high} | Significant security concern |`);
+    }
+    if (by_severity.medium > 0) {
+      lines.push(`| Medium | ${by_severity.medium} | Moderate security issue |`);
+    }
+    if (by_severity.low > 0) {
+      lines.push(`| Low | ${by_severity.low} | Minor security concern |`);
+    }
+    lines.push("");
+    lines.push("---");
+    lines.push("");
+  }
+  const unsafeResults = manifest.results.filter((r) => r.status === "unsafe");
+  if (unsafeResults.length > 0) {
+    lines.push("## Vulnerabilities Found");
+    lines.push("");
+    for (let i = 0;i < unsafeResults.length; i++) {
+      const result = unsafeResults[i];
+      lines.push(`### ${i + 1}. ${result.mutation} (${result.severity.toUpperCase()})`);
+      lines.push("");
+      lines.push(`**Case ID:** ${result.caseId}`);
+      lines.push(`**Severity:** ${result.severity.toUpperCase()}`);
+      lines.push(`**Mutation:** ${result.mutation}`);
+      lines.push("");
+      if (includeDetails) {
+        lines.push("**Attack Prompt:**");
+        lines.push("```");
+        lines.push(truncate(result.prompt, truncateAt));
+        lines.push("```");
+        lines.push("");
+        lines.push("**Model Response:**");
+        lines.push("```");
+        lines.push(truncate(result.response || "(empty)", truncateAt));
+        lines.push("```");
+        lines.push("");
+      }
+      if (result.reasons.length > 0) {
+        lines.push("**Reasons:**");
+        for (const reason of result.reasons) {
+          lines.push(`- ${reason}`);
+        }
+        lines.push("");
+      }
+      lines.push("---");
+      lines.push("");
+    }
+  }
+  lines.push("## Recommendations");
+  lines.push("");
+  if (manifest.metrics.unsafe_responses === 0) {
+    lines.push("- All attacks were successfully defended");
+    lines.push("- Continue monitoring and testing regularly");
+  } else {
+    if (by_severity.critical > 0 || by_severity.high > 0) {
+      lines.push("- **High Priority:** Review and address critical/high severity vulnerabilities immediately");
+    }
+    if (manifest.metrics.unsafe_responses > testable * 0.1) {
+      lines.push("- Consider implementing additional input validation and output filtering");
+    }
+    lines.push("- Review system prompts for potential information leakage");
+    lines.push("- Implement role-play and persona detection");
+    lines.push("- Add output filtering for dangerous content patterns");
+  }
+  lines.push("");
+  lines.push("---");
+  lines.push("");
+  lines.push(`*Generated by [ArtemisKit](https://artemiskit.vercel.app) v${manifest.version}*`);
+  return lines.join(`
+`);
+}
 export {
   generateStressHTMLReport,
+  generateRedTeamMarkdownReport,
   generateRedTeamHTMLReport,
+  generateMarkdownReport,
   generateJSONReport,
   generateHTMLReport,
   generateCompareHTMLReport,

package/dist/markdown/generator.d.ts

@@ -0,0 +1,21 @@
+/**
+ * Markdown Report Generator
+ *
+ * Generates documentation-friendly markdown reports for compliance and audit trails.
+ */
+import type { RedTeamManifest, RunManifest } from '@artemiskit/core';
+export interface MarkdownReportOptions {
+    /** Include full prompt/response details for failed cases */
+    includeDetails?: boolean;
+    /** Maximum characters to show for prompts/responses */
+    truncateAt?: number;
+}
+/**
+ * Generate markdown report for a standard run
+ */
+export declare function generateMarkdownReport(manifest: RunManifest, options?: MarkdownReportOptions): string;
+/**
+ * Generate markdown report for red team results
+ */
+export declare function generateRedTeamMarkdownReport(manifest: RedTeamManifest, options?: MarkdownReportOptions): string;
+//# sourceMappingURL=generator.d.ts.map

package/dist/markdown/generator.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"generator.d.ts","sourceRoot":"","sources":["../../src/markdown/generator.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,OAAO,KAAK,EAAE,eAAe,EAAE,WAAW,EAAE,MAAM,kBAAkB,CAAC;AAErE,MAAM,WAAW,qBAAqB;IACpC,4DAA4D;IAC5D,cAAc,CAAC,EAAE,OAAO,CAAC;IACzB,uDAAuD;IACvD,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAkCD;;GAEG;AACH,wBAAgB,sBAAsB,CACpC,QAAQ,EAAE,WAAW,EACrB,OAAO,GAAE,qBAA0B,GAClC,MAAM,CAuJR;AAED;;GAEG;AACH,wBAAgB,6BAA6B,CAC3C,QAAQ,EAAE,eAAe,EACzB,OAAO,GAAE,qBAA0B,GAClC,MAAM,CAmJR"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@artemiskit/reports",
-  "version": "0.2.2",
+  "version": "0.2.3",
   "description": "HTML report generation for ArtemisKit LLM evaluation toolkit",
   "type": "module",
   "license": "Apache-2.0",
@@ -30,7 +30,7 @@
     "test": "bun test"
   },
   "dependencies": {
-    "@artemiskit/core": "workspace:*",
+    "@artemiskit/core": "0.2.3",
     "handlebars": "^4.7.8"
   },
   "devDependencies": {

package/src/index.ts

@@ -20,3 +20,10 @@ export {
   type ComparisonData,
   type CaseComparison,
 } from './html/compare-generator';
+
+// Markdown reports
+export {
+  generateMarkdownReport,
+  generateRedTeamMarkdownReport,
+  type MarkdownReportOptions,
+} from './markdown/generator';

package/src/markdown/generator.ts

@@ -0,0 +1,360 @@
+/**
+ * Markdown Report Generator
+ *
+ * Generates documentation-friendly markdown reports for compliance and audit trails.
+ */
+
+import type { RedTeamManifest, RunManifest } from '@artemiskit/core';
+
+export interface MarkdownReportOptions {
+  /** Include full prompt/response details for failed cases */
+  includeDetails?: boolean;
+  /** Maximum characters to show for prompts/responses */
+  truncateAt?: number;
+}
+
+/**
+ * Truncate text to a maximum length
+ */
+function truncate(text: string, maxLength: number): string {
+  if (text.length <= maxLength) return text;
+  return `${text.slice(0, maxLength)}...`;
+}
+
+/**
+ * Format cost for display
+ */
+function formatCostMd(costUsd: number): string {
+  if (costUsd < 0.01) {
+    return `$${(costUsd * 100).toFixed(4)} cents`;
+  }
+  if (costUsd < 1) {
+    return `$${costUsd.toFixed(4)}`;
+  }
+  return `$${costUsd.toFixed(2)}`;
+}
+
+/**
+ * Format duration in milliseconds to human-readable string
+ */
+function formatDuration(ms: number): string {
+  if (ms < 1000) return `${ms}ms`;
+  if (ms < 60000) return `${(ms / 1000).toFixed(1)}s`;
+  const minutes = Math.floor(ms / 60000);
+  const seconds = ((ms % 60000) / 1000).toFixed(0);
+  return `${minutes}m ${seconds}s`;
+}
+
+/**
+ * Generate markdown report for a standard run
+ */
+export function generateMarkdownReport(
+  manifest: RunManifest,
+  options: MarkdownReportOptions = {}
+): string {
+  const { includeDetails = true, truncateAt = 500 } = options;
+  const lines: string[] = [];
+
+  // Header
+  lines.push('# ArtemisKit Test Results');
+  lines.push('');
+  lines.push(`**Scenario:** ${manifest.config.scenario}`);
+  lines.push(`**Run ID:** ${manifest.run_id}`);
+  lines.push(`**Date:** ${new Date(manifest.start_time).toISOString()}`);
+  lines.push(
+    `**Provider:** ${manifest.config.provider}${manifest.config.model ? ` (${manifest.config.model})` : ''}`
+  );
+  lines.push('');
+  lines.push('---');
+  lines.push('');
+
+  // Summary table
+  lines.push('## Summary');
+  lines.push('');
+  lines.push('| Metric | Value |');
+  lines.push('|--------|-------|');
+  lines.push(`| Total Cases | ${manifest.metrics.total_cases} |`);
+  lines.push(
+    `| Passed | ${manifest.metrics.passed_cases} (${(manifest.metrics.success_rate * 100).toFixed(1)}%) |`
+  );
+  lines.push(`| Failed | ${manifest.metrics.failed_cases} |`);
+  lines.push(`| Duration | ${formatDuration(manifest.duration_ms)} |`);
+  lines.push(`| Median Latency | ${manifest.metrics.median_latency_ms}ms |`);
+  lines.push(`| P95 Latency | ${manifest.metrics.p95_latency_ms}ms |`);
+  lines.push(`| Total Tokens | ${manifest.metrics.total_tokens.toLocaleString()} |`);
+
+  if (manifest.metrics.cost) {
+    lines.push(`| Estimated Cost | ${formatCostMd(manifest.metrics.cost.total_usd)} |`);
+  }
+
+  lines.push('');
+  lines.push('---');
+  lines.push('');
+
+  // Results by case
+  lines.push('## Results by Case');
+  lines.push('');
+
+  // Passed cases (collapsed)
+  const passed = manifest.cases.filter((c) => c.ok);
+  lines.push(`### Passed (${passed.length})`);
+  lines.push('');
+
+  if (passed.length > 0) {
+    lines.push('<details>');
+    lines.push('<summary>Click to expand passed cases</summary>');
+    lines.push('');
+    lines.push('| Case ID | Latency | Tokens | Score |');
+    lines.push('|---------|---------|--------|-------|');
+    for (const c of passed) {
+      lines.push(
+        `| ${c.id} | ${formatDuration(c.latencyMs)} | ${c.tokens?.total || '-'} | ${(c.score * 100).toFixed(0)}% |`
+      );
+    }
+    lines.push('');
+    lines.push('</details>');
+  } else {
+    lines.push('_No passed cases_');
+  }
+
+  lines.push('');
+
+  // Failed cases (expanded with details)
+  const failed = manifest.cases.filter((c) => !c.ok);
+  lines.push(`### Failed (${failed.length})`);
+  lines.push('');
+
+  if (failed.length > 0) {
+    for (const c of failed) {
+      lines.push(`#### \`${c.id}\``);
+      lines.push('');
+
+      if (includeDetails) {
+        // Prompt
+        const promptStr =
+          typeof c.prompt === 'string' ? c.prompt : JSON.stringify(c.prompt, null, 2);
+        lines.push('**Prompt:**');
+        lines.push('```');
+        lines.push(truncate(promptStr, truncateAt));
+        lines.push('```');
+        lines.push('');
+
+        // Expected
+        lines.push('**Expected:**');
+        lines.push(`- Type: \`${c.matcherType}\``);
+        lines.push('```json');
+        lines.push(truncate(JSON.stringify(c.expected, null, 2), truncateAt));
+        lines.push('```');
+        lines.push('');
+
+        // Actual response
+        lines.push('**Actual Response:**');
+        lines.push('```');
+        lines.push(truncate(c.response || '(empty)', truncateAt));
+        lines.push('```');
+        lines.push('');
+      }
+
+      // Reason
+      lines.push(`**Reason:** ${c.reason || 'Unknown'}`);
+      lines.push('');
+      lines.push('---');
+      lines.push('');
+    }
+  } else {
+    lines.push('_No failed cases_');
+    lines.push('');
+  }
+
+  // Configuration section
+  if (manifest.resolved_config) {
+    lines.push('## Configuration');
+    lines.push('');
+    lines.push('```yaml');
+    lines.push(`provider: ${manifest.resolved_config.provider}`);
+    if (manifest.resolved_config.model) {
+      lines.push(`model: ${manifest.resolved_config.model}`);
+    }
+    if (manifest.resolved_config.temperature !== undefined) {
+      lines.push(`temperature: ${manifest.resolved_config.temperature}`);
+    }
+    if (manifest.resolved_config.max_tokens !== undefined) {
+      lines.push(`max_tokens: ${manifest.resolved_config.max_tokens}`);
+    }
+    lines.push('```');
+    lines.push('');
+  }
+
+  // Redaction info
+  if (manifest.redaction?.enabled) {
+    lines.push('## Redaction');
+    lines.push('');
+    lines.push(`- **Patterns Used:** ${manifest.redaction.patternsUsed.join(', ')}`);
+    lines.push(`- **Prompts Redacted:** ${manifest.redaction.summary.promptsRedacted}`);
+    lines.push(`- **Responses Redacted:** ${manifest.redaction.summary.responsesRedacted}`);
+    lines.push(`- **Total Redactions:** ${manifest.redaction.summary.totalRedactions}`);
+    lines.push('');
+  }
+
+  // Footer
+  lines.push('---');
+  lines.push('');
+  lines.push(`*Generated by [ArtemisKit](https://artemiskit.vercel.app) v${manifest.version}*`);
+
+  return lines.join('\n');
+}
+
+/**
+ * Generate markdown report for red team results
+ */
+export function generateRedTeamMarkdownReport(
+  manifest: RedTeamManifest,
+  options: MarkdownReportOptions = {}
+): string {
+  const { includeDetails = true, truncateAt = 500 } = options;
+  const lines: string[] = [];
+
+  // Header
+  lines.push('# ArtemisKit Security Report');
+  lines.push('');
+  lines.push(`**Scenario:** ${manifest.config.scenario}`);
+  lines.push(`**Run ID:** ${manifest.run_id}`);
+  lines.push(`**Date:** ${new Date(manifest.start_time).toISOString()}`);
+  lines.push('**Test Type:** Red Team Security Scan');
+  lines.push('');
+  lines.push('---');
+  lines.push('');
+
+  // Security Summary
+  const testable = manifest.metrics.total_tests - manifest.metrics.error_responses;
+  const defenseRate = manifest.metrics.defense_rate * 100;
+
+  // Determine risk level
+  let riskLevel: string;
+  let riskEmoji: string;
+  if (defenseRate >= 95) {
+    riskLevel = 'LOW';
+    riskEmoji = '';
+  } else if (defenseRate >= 80) {
+    riskLevel = 'MEDIUM';
+    riskEmoji = '';
+  } else if (defenseRate >= 50) {
+    riskLevel = 'HIGH';
+    riskEmoji = '';
+  } else {
+    riskLevel = 'CRITICAL';
+    riskEmoji = '';
+  }
+
+  lines.push('## Security Summary');
+  lines.push('');
+  lines.push('| Metric | Value |');
+  lines.push('|--------|-------|');
+  lines.push(`| Defense Rate | ${defenseRate.toFixed(1)}% |`);
+  lines.push(`| Total Attacks | ${manifest.metrics.total_tests} |`);
+  lines.push(`| Defended | ${manifest.metrics.defended} |`);
+  lines.push(`| Safe Responses | ${manifest.metrics.safe_responses} |`);
+  lines.push(`| Blocked | ${manifest.metrics.blocked_responses} |`);
+  lines.push(`| Vulnerabilities | ${manifest.metrics.unsafe_responses} |`);
+  lines.push(`| Errors | ${manifest.metrics.error_responses} |`);
+  lines.push(`| Risk Level | **${riskEmoji} ${riskLevel}** |`);
+  lines.push('');
+  lines.push('---');
+  lines.push('');
+
+  // Severity Breakdown
+  const { by_severity } = manifest.metrics;
+  if (manifest.metrics.unsafe_responses > 0) {
+    lines.push('## Severity Breakdown');
+    lines.push('');
+    lines.push('| Severity | Count | Description |');
+    lines.push('|----------|-------|-------------|');
+    if (by_severity.critical > 0) {
+      lines.push(`| Critical | ${by_severity.critical} | Severe security vulnerability |`);
+    }
+    if (by_severity.high > 0) {
+      lines.push(`| High | ${by_severity.high} | Significant security concern |`);
+    }
+    if (by_severity.medium > 0) {
+      lines.push(`| Medium | ${by_severity.medium} | Moderate security issue |`);
+    }
+    if (by_severity.low > 0) {
+      lines.push(`| Low | ${by_severity.low} | Minor security concern |`);
+    }
+    lines.push('');
+    lines.push('---');
+    lines.push('');
+  }
+
+  // Vulnerabilities Found
+  const unsafeResults = manifest.results.filter((r) => r.status === 'unsafe');
+  if (unsafeResults.length > 0) {
+    lines.push('## Vulnerabilities Found');
+    lines.push('');
+
+    for (let i = 0; i < unsafeResults.length; i++) {
+      const result = unsafeResults[i];
+      lines.push(`### ${i + 1}. ${result.mutation} (${result.severity.toUpperCase()})`);
+      lines.push('');
+      lines.push(`**Case ID:** ${result.caseId}`);
+      lines.push(`**Severity:** ${result.severity.toUpperCase()}`);
+      lines.push(`**Mutation:** ${result.mutation}`);
+      lines.push('');
+
+      if (includeDetails) {
+        lines.push('**Attack Prompt:**');
+        lines.push('```');
+        lines.push(truncate(result.prompt, truncateAt));
+        lines.push('```');
+        lines.push('');
+
+        lines.push('**Model Response:**');
+        lines.push('```');
+        lines.push(truncate(result.response || '(empty)', truncateAt));
+        lines.push('```');
+        lines.push('');
+      }
+
+      if (result.reasons.length > 0) {
+        lines.push('**Reasons:**');
+        for (const reason of result.reasons) {
+          lines.push(`- ${reason}`);
+        }
+        lines.push('');
+      }
+
+      lines.push('---');
+      lines.push('');
+    }
+  }
+
+  // Recommendations
+  lines.push('## Recommendations');
+  lines.push('');
+
+  if (manifest.metrics.unsafe_responses === 0) {
+    lines.push('- All attacks were successfully defended');
+    lines.push('- Continue monitoring and testing regularly');
+  } else {
+    if (by_severity.critical > 0 || by_severity.high > 0) {
+      lines.push(
+        '- **High Priority:** Review and address critical/high severity vulnerabilities immediately'
+      );
+    }
+    if (manifest.metrics.unsafe_responses > testable * 0.1) {
+      lines.push('- Consider implementing additional input validation and output filtering');
+    }
+    lines.push('- Review system prompts for potential information leakage');
+    lines.push('- Implement role-play and persona detection');
+    lines.push('- Add output filtering for dangerous content patterns');
+  }
+
+  lines.push('');
+
+  // Footer
+  lines.push('---');
+  lines.push('');
+  lines.push(`*Generated by [ArtemisKit](https://artemiskit.vercel.app) v${manifest.version}*`);
+
+  return lines.join('\n');
+}