torusguard 1.3.4 → 1.3.5

package/.torusguard/scripts/harden_runner.py

@@ -1,455 +1,628 @@
-#!/usr/bin/env python3
-"""
-TorusGuard Autonomous Remediation Engine (v1.3.3)
-Evaluates audit findings against canonical patch templates, enforces Ponytail bounds
-(<= 35 additions, <= 25 deletions), packages structured candidate remediation bundles,
-and renders unified patch diffs with standardized 75-column terminal UI.
-
-Pure Python 3.10+ standard library (zero external dependencies).
-"""
-
-import sys
-import os
-import re
-import json
-import difflib
-import hashlib
-import argparse
-import datetime
-from pathlib import Path
-from typing import Dict, List, Any, Optional
-
-# Ensure scripts dir is in sys.path for term_ui import
-scripts_dir = Path(__file__).resolve().parent
-if str(scripts_dir) not in sys.path:
-    sys.path.insert(0, str(scripts_dir))
-
-try:
-    import term_ui
-except ImportError:
-    term_ui = None
-
-# Windows console UTF-8 support
-if sys.stdout and hasattr(sys.stdout, "reconfigure"):
-    try:
-        getattr(sys.stdout, "reconfigure")(encoding="utf-8", errors="replace")
-    except Exception:
-        pass
-
-# ─── Timezone Configuration: Indian Standard Time (IST, UTC+05:30) ───────────
-IST = datetime.timezone(datetime.timedelta(hours=5, minutes=30), name="IST")
-
-def get_ist_now() -> datetime.datetime:
-    return datetime.datetime.now(IST)
-
-# ─── Fallback Formatters (if term_ui missing) ──────────────────────────────────
-BOLD = "\033[1m"
-DIM = "\033[2m"
-RESET = "\033[0m"
-CYAN = "\033[36m"
-GREEN = "\033[32m"
-YELLOW = "\033[33m"
-WHITE = "\033[97m"
-GRAY = "\033[90m"
-RED = "\033[31m"
-
-def box_line(content: str, width: int = 67, border: str = "│", border_color: str = CYAN) -> str:
-    if term_ui:
-        return term_ui.format_box_line(content, width=width, border=border, border_color=border_color)
-    return f"  {border_color}{border}{RESET}  {content}"
-
-def box_header(title: str, subtitle: str = "", version: str = "v1.3.3", border_color: str = CYAN) -> str:
-    if term_ui:
-        return term_ui.card_header(title, subtitle, version, border_color)
-    return f"=== {title} ({version}) ==="
-
-def border_top(title: str = "", border_color: str = CYAN, double: bool = False) -> str:
-    if term_ui:
-        return term_ui.card_border_top(title, border_color=border_color, double=double)
-    return "┌" + "─" * 71 + "┐"
-
-def border_bottom(border_color: str = CYAN, double: bool = False) -> str:
-    if term_ui:
-        return term_ui.card_border_bottom(border_color=border_color, double=double)
-    return "└" + "─" * 71 + "┘"
-
-
-def find_latest_audit_run(runs_dir: Path) -> Optional[Path]:
-    """Locate the most recent audit run folder containing findings.json."""
-    if not runs_dir.is_dir():
-        return None
-    runs = sorted(runs_dir.glob("run-*-audit"), key=lambda p: p.stat().st_mtime, reverse=True)
-    for r in runs:
-        if (r / "findings.json").is_file():
-            return r
-    return None
-
-
-def generate_patch_for_finding(finding: Dict[str, Any], target_root: Path) -> Optional[Dict[str, Any]]:
-    """
-    Formulate a minimal, surgical fix for a finding adhering to Ponytail bounds
-    (<= 35 additions, <= 25 deletions).
-    """
-    rule_id = finding.get("rule_id", "")
-    file_path = finding.get("file_path", "")
-    line_number = finding.get("line_number", 1)
-    description = finding.get("description", "")
-
-    full_target = target_root / file_path
-    if not full_target.is_file():
-        return None
-
-    try:
-        content = full_target.read_text(encoding="utf-8", errors="replace")
-    except Exception:
-        return None
-
-    lines = content.splitlines()
-    if not (1 <= line_number <= len(lines)):
-        return None
-
-    target_line = lines[line_number - 1]
-    indent = re.match(r'^\s*', target_line).group(0)  # type: ignore
-    new_lines = list(lines)
-    applied_rule = False
-    fix_explanation = ""
-
-    # Strategy 1: Hardcoded Secrets (TG-SEC-001 / TG-SEC-002)
-    if rule_id in ("TG-SEC-001", "TG-SEC-002"):
-        if file_path.endswith((".js", ".jsx", ".ts", ".tsx", ".mjs")):
-            if re.search(r'(jwt_secret|jwtSecret|JWT_SECRET)', target_line, re.IGNORECASE):
-                var_match = re.search(r'(const|let|var)\s+([A-Za-z0-9_]+)\s*=', target_line)
-                var_name = var_match.group(2) if var_match else "JWT_SECRET"
-                replacement = f'{indent}const {var_name} = process.env.JWT_SECRET || "";'
-                new_lines[line_number - 1] = replacement
-                applied_rule = True
-                fix_explanation = f"Migrated hardcoded JWT secret to environment variable process.env.JWT_SECRET"
-            elif re.search(r'(:=|=)\s*["\'][A-Za-z0-9_\-]{8,}["\']', target_line):
-                var_match = re.search(r'(const|let|var)?\s*([A-Za-z0-9_]+)\s*[:=]', target_line)
-                var_name = var_match.group(2) if var_match else "API_KEY"
-                env_key = re.sub(r'(?<!^)(?=[A-Z])', '_', var_name).upper()
-                replacement = f'{indent}const {var_name} = process.env.{env_key} || "";'
-                new_lines[line_number - 1] = replacement
-                applied_rule = True
-                fix_explanation = f"Replaced hardcoded credential with process.env.{env_key}"
-        elif file_path.endswith(".py"):
-            if re.search(r'(:=|=)\s*["\'][A-Za-z0-9_\-]{8,}["\']', target_line):
-                var_match = re.search(r'([A-Za-z0-9_]+)\s*=', target_line)
-                var_name = var_match.group(1) if var_match else "SECRET_KEY"
-                env_key = var_name.upper()
-                replacement = f'{indent}{var_name} = os.environ.get("{env_key}", "")'
-                new_lines[line_number - 1] = replacement
-                applied_rule = True
-                fix_explanation = f"Replaced hardcoded credential with os.environ.get('{env_key}')"
-
-    # Strategy 2: Dangerous React HTML rendering & DOM innerHTML (TG-INPUT-003)
-    elif rule_id == "TG-INPUT-003":
-        if "dangerouslySetInnerHTML" in target_line:
-            var_match = re.search(r'__html\s*:\s*([A-Za-z0-9_$.]+)', target_line)
-            var_name = var_match.group(1) if var_match else "content"
-            replacement = f'{indent}<div>{{{var_name}}}</div>'
-            new_lines[line_number - 1] = replacement
-            applied_rule = True
-            fix_explanation = "Replaced raw dangerouslySetInnerHTML injection with safe React text interpolation"
-        elif ".innerHTML" in target_line:
-            if re.search(r'\.innerHTML\s*=\s*["\']["\']', target_line):
-                fixed_line = re.sub(r'([a-zA-Z0-9_$]+)\.innerHTML\s*=\s*["\']["\']', r'\1.textContent = ""', target_line)
-                if fixed_line != target_line:
-                    new_lines[line_number - 1] = fixed_line
-                    applied_rule = True
-                    fix_explanation = "Replaced unsafe innerHTML DOM reset with safe textContent assignment"
-            elif re.search(r'([a-zA-Z0-9_$]+)\.innerHTML\s*=\s*([a-zA-Z0-9_$.]+)\s*;?$', target_line):
-                fixed_line = re.sub(r'([a-zA-Z0-9_$]+)\.innerHTML\s*=\s*([a-zA-Z0-9_$.]+)', r'\1.textContent = \2', target_line)
-                if fixed_line != target_line:
-                    new_lines[line_number - 1] = fixed_line
-                    applied_rule = True
-                    fix_explanation = "Replaced raw DOM innerHTML assignment with XSS-safe textContent"
-
-    # Strategy 3: Raw SQL Concatenation (TG-INPUT-002)
-    elif rule_id == "TG-INPUT-002":
-        if file_path.endswith(".py") and ("f\"SELECT" in target_line or "f'SELECT" in target_line):
-            fixed_line = re.sub(r'f(["\']SELECT.*?WHERE.*?)(\{[^}]+\})', r'\1%s', target_line)
-            if fixed_line != target_line:
-                new_lines[line_number - 1] = fixed_line
-                applied_rule = True
-                fix_explanation = "Replaced unsafe Python SQL f-string interpolation with parameterized SQL query placeholder"
-        elif file_path.endswith((".js", ".jsx", ".ts", ".tsx", ".mjs")):
-            # Template string SQL injection: `SELECT ... ${var}` -> parameterized placeholder
-            if "`SELECT" in target_line or "`select" in target_line:
-                fixed_line = re.sub(r'\$\{[^}]+\}', '?', target_line)
-                if fixed_line != target_line:
-                    new_lines[line_number - 1] = fixed_line
-                    applied_rule = True
-                    fix_explanation = "Replaced template literal SQL string interpolation with parameterized SQL placeholder (?)"
-
-    # Strategy 4: Multi-Tenant Query Scoping (TG-DB-004)
-    elif rule_id == "TG-DB-004":
-        if ".objects.get(" in target_line and "tenant" not in target_line.lower() and "org" not in target_line.lower():
-            fixed_line = target_line.replace(".objects.get(", ".objects.get(organization_id=request.user.organization_id, ")
-            if fixed_line != target_line:
-                new_lines[line_number - 1] = fixed_line
-                applied_rule = True
-                fix_explanation = "Added explicit organization_id tenant boundary filter to ORM query"
-        elif "findUnique(" in target_line and "tenant" not in target_line.lower():
-            fixed_line = target_line.replace("findUnique({ where: {", "findFirst({ where: { tenantId: req.tenantId,")
-            if fixed_line != target_line:
-                new_lines[line_number - 1] = fixed_line
-                applied_rule = True
-                fix_explanation = "Enforced tenantId multi-tenant isolation in Prisma lookup"
-
-    # Strategy 5: Permissive CORS with Credentials (TG-PLATFORM-001)
-    elif rule_id == "TG-PLATFORM-001":
-        if "CORS_ALLOW_ALL_ORIGINS = True" in target_line:
-            replacement = f'{indent}CORS_ALLOWED_ORIGINS = [os.environ.get("ALLOWED_ORIGIN", "https://app.example.com")]'
-            new_lines[line_number - 1] = replacement
-            applied_rule = True
-            fix_explanation = "Replaced wildcard CORS configuration with strict origin whitelist environment variable"
-        elif "cors(" in target_line and ("origin: '*'" in target_line or 'origin: "*"' in target_line):
-            fixed_line = target_line.replace("origin: '*'", "origin: process.env.ALLOWED_ORIGIN || 'http://localhost:5173'")
-            fixed_line = fixed_line.replace('origin: "*"', "origin: process.env.ALLOWED_ORIGIN || 'http://localhost:5173'")
-            if fixed_line != target_line:
-                new_lines[line_number - 1] = fixed_line
-                applied_rule = True
-                fix_explanation = "Constrained wildcard CORS origin to process.env.ALLOWED_ORIGIN with secure fallback"
-
-    # Strategy 6: Cookie Missing Security Flags (TG-AUTH-004)
-    elif rule_id == "TG-AUTH-004":
-        if "res.cookie(" in target_line and "httpOnly" not in target_line:
-            # res.cookie('token', token) -> res.cookie('token', token, { httpOnly: true, secure: process.env.NODE_ENV === 'production', sameSite: 'lax' })
-            fixed_line = re.sub(
-                r'res\.cookie\(\s*([a-zA-Z0-9_\'\"]+)\s*,\s*([a-zA-Z0-9_]+)\s*\)',
-                r"res.cookie(\1, \2, { httpOnly: true, secure: process.env.NODE_ENV === 'production', sameSite: 'lax' })",
-                target_line
-            )
-            if fixed_line != target_line:
-                new_lines[line_number - 1] = fixed_line
-                applied_rule = True
-                fix_explanation = "Injected mandatory security flags (httpOnly, secure, sameSite) to cookie definition"
-
-    # Strategy 7: Path Traversal (TG-INPUT-006)
-    elif rule_id == "TG-INPUT-006":
-        if "path.join(" in target_line and "path.basename" not in target_line:
-            fixed_line = re.sub(r'path\.join\((.*?),\s*([a-zA-Z0-9_$.]+)\)', r'path.join(\1, path.basename(\2))', target_line)
-            if fixed_line != target_line:
-                new_lines[line_number - 1] = fixed_line
-                applied_rule = True
-                fix_explanation = "Wrapped dynamic path segment in path.basename to neutralize path traversal sequences"
-        elif "os.path.join(" in target_line and "os.path.basename" not in target_line:
-            fixed_line = re.sub(r'os\.path\.join\((.*?),\s*([a-zA-Z0-9_$.]+)\)', r'os.path.join(\1, os.path.basename(\2))', target_line)
-            if fixed_line != target_line:
-                new_lines[line_number - 1] = fixed_line
-                applied_rule = True
-                fix_explanation = "Wrapped dynamic path segment in os.path.basename to neutralize path traversal sequences"
-
-    # Strategy 8: TLS Verification Bypass (TG-DIFF-001)
-    elif rule_id == "TG-DIFF-001":
-        if "verify=False" in target_line:
-            fixed_line = target_line.replace("verify=False", "verify=True")
-            new_lines[line_number - 1] = fixed_line
-            applied_rule = True
-            fix_explanation = "Restored mandatory TLS certificate verification (verify=True)"
-        elif "rejectUnauthorized: false" in target_line:
-            fixed_line = target_line.replace("rejectUnauthorized: false", "rejectUnauthorized: true")
-            new_lines[line_number - 1] = fixed_line
-            applied_rule = True
-            fix_explanation = "Restored TLS certificate authority validation (rejectUnauthorized: true)"
-        elif "InsecureSkipVerify: true" in target_line:
-            fixed_line = target_line.replace("InsecureSkipVerify: true", "InsecureSkipVerify: false")
-            new_lines[line_number - 1] = fixed_line
-            applied_rule = True
-            fix_explanation = "Restored mandatory TLS certificate validation (InsecureSkipVerify: false)"
-
-    if not applied_rule:
-        return None
-
-    # Calculate unified diff
-    orig_lines = [l + "\n" for l in lines]
-    patched_lines = [l + "\n" for l in new_lines]
-    diff = "".join(difflib.unified_diff(
-        orig_lines, patched_lines,
-        fromfile=f"a/{file_path}", tofile=f"b/{file_path}",
-        n=2
-    ))
-
-    additions = sum(1 for l in diff.splitlines() if l.startswith("+") and not l.startswith("+++"))
-    deletions = sum(1 for l in diff.splitlines() if l.startswith("-") and not l.startswith("---"))
-
-    # Ponytail bounds validation: <= 35 additions, <= 25 deletions
-    if additions > 35 or deletions > 25:
-        return None
-
-    bundle_id = f"bnd-{rule_id.lower()}-{line_number}-{hashlib.sha256(diff.encode('utf-8')).hexdigest()[:6]}"
-
-    return {
-        "bundle_id": bundle_id,
-        "finding_id": finding.get("finding_id", f"{rule_id}-{line_number}"),
-        "rule_id": rule_id,
-        "title": finding.get("title", rule_id),
-        "target_file": file_path,
-        "line_number": line_number,
-        "target_line": target_line,
-        "replacement_line": new_lines[line_number - 1],
-        "what_is_wrong": description,
-        "why_it_matters": "Security vulnerability violating TorusGuard strict production safety invariant.",
-        "what_should_change": fix_explanation,
-        "proposed_diff": diff,
-        "additions": additions,
-        "deletions": deletions,
-        "patched_content": "\n".join(new_lines)
-    }
-
-
-def execute_harden(target_root: Path, run_id_arg: Optional[str] = None) -> Dict[str, Any]:
-    """Execute autonomous remediation formulation and package candidate bundles."""
-    runs_dir = target_root / ".torusguard" / "runs"
-    if run_id_arg:
-        run_folder = runs_dir / run_id_arg
-    else:
-        run_folder = find_latest_audit_run(runs_dir)
-
-    if not run_folder or not run_folder.is_dir():
-        print(f"\n  {RED}✖ No audit runs found to harden.{RESET}")
-        print(f"  Run {CYAN}npx torusguard audit{RESET} first to discover actionable findings.\n")
-        sys.exit(1)
-
-    findings_file = run_folder / "findings.json"
-    if not findings_file.is_file():
-        print(f"\n  {RED}✖ Missing findings.json in run folder:{RESET} {run_folder}\n")
-        sys.exit(1)
-
-    try:
-        findings = json.loads(findings_file.read_text(encoding="utf-8"))
-    except Exception as e:
-        print(f"\n  {RED}✖ Failed to parse findings.json:{RESET} {e}\n")
-        sys.exit(1)
-
-    bundles_dir = run_folder / "bundles"
-    bundles_dir.mkdir(parents=True, exist_ok=True)
-
-    bundles = []
-    seen_file_lines = set()
-    for f in findings:
-        target_key = (f.get("file_path"), f.get("line_number"))
-        if target_key in seen_file_lines:
-            continue
-        candidate = generate_patch_for_finding(f, target_root)
-        if candidate:
-            seen_file_lines.add(target_key)
-            bundles.append(candidate)
-            b_dir = bundles_dir / candidate["bundle_id"]
-            b_dir.mkdir(parents=True, exist_ok=True)
-
-            # 1. patch.diff
-            (b_dir / "patch.diff").write_text(candidate["proposed_diff"], encoding="utf-8")
-
-            # 2. minimal_patch_plan.md
-            plan_content = [
-                f"# Remediation Plan: {candidate['bundle_id']}",
-                f"- **Rule ID:** `{candidate['rule_id']}`",
-                f"- **Target File:** `{candidate['target_file']}:{candidate['line_number']}`",
-                f"- **Ponytail Churn:** `+{candidate['additions']} / -{candidate['deletions']}` (Compliant <=35/<=25)",
-                f"\n## Proposed Change\n{candidate['what_should_change']}\n",
-                "## Unified Diff Preview\n```diff",
-                candidate["proposed_diff"],
-                "```"
-            ]
-            (b_dir / "minimal_patch_plan.md").write_text("\n".join(plan_content), encoding="utf-8")
-
-            # 3. metadata.json
-            with open(b_dir / "metadata.json", "w", encoding="utf-8") as mf:
-                json.dump(candidate, mf, indent=2)
-
-    # Render run-level remediation.md
-    remediation_lines = [
-        f"# TorusGuard Governed Remediation Catalog",
-        f"**Run ID:** `{run_folder.name}`  ",
-        f"**Generated:** `{get_ist_now().strftime('%Y-%m-%d %H:%M:%S IST')}`  ",
-        f"**Candidate Bundles Formulated:** `{len(bundles)}`  \n",
-        "---",
-        "## Formulated Candidate Patches\n"
-    ]
-
-    for b in bundles:
-        remediation_lines.append(f"### [{b['rule_id']}] {b['title']}")
-        remediation_lines.append(f"- **Target:** `{b['target_file']}:{b['line_number']}` | **Bundle ID:** `{b['bundle_id']}`")
-        remediation_lines.append(f"- **Strategy:** {b['what_should_change']}")
-        remediation_lines.append(f"- **Ponytail Churn:** `+{b['additions']} / -{b['deletions']}`")
-        remediation_lines.append("```diff")
-        remediation_lines.append(b["proposed_diff"].strip())
-        remediation_lines.append("```\n")
-
-    (run_folder / "remediation.md").write_text("\n".join(remediation_lines), encoding="utf-8")
-
-    # Update manifest
-    manifest_file = run_folder / "manifest.json"
-    if manifest_file.is_file():
-        try:
-            mdata = json.loads(manifest_file.read_text(encoding="utf-8"))
-            mdata["remediation_bundles_count"] = len(bundles)
-            manifest_file.write_text(json.dumps(mdata, indent=2), encoding="utf-8")
-        except Exception:
-            pass
-
-    # Print 75-column Terminal Cards
-    print()
-    print(box_header("🛡️  TORUSGUARD GOVERNED REMEDIATION ENGINE", "Autonomous Minimal Patch Formulation & Ponytail Packaging", "v1.3.3"))
-    print()
-
-    print(border_top("Remediation Scope"))
-    print(box_line(f"Active Run:     {WHITE}{run_folder.name}{RESET}"))
-    print(box_line(f"Target Scope:   {WHITE}{target_root}{RESET}"))
-    print(box_line(f"Total Findings: {len(findings)} evaluated | {GREEN}{len(bundles)} surgical patches formulated{RESET}"))
-    print(box_line(f"Ponytail Guard: {GREEN}All patches strictly <= 35 additions, <= 25 deletions{RESET}"))
-    print(border_bottom())
-    print()
-
-    if bundles:
-        print(border_top(f"Formulated Candidate Patches ({len(bundles)})"))
-        for b in bundles[:5]:
-            print(box_line(f"{YELLOW}[{b['rule_id']}]{RESET} {WHITE}{b['target_file']}:{b['line_number']}{RESET} (+{b['additions']}/-{b['deletions']})"))
-            print(box_line(f"  └─ {DIM}{b['what_should_change']}{RESET}"))
-        if len(bundles) > 5:
-            print(box_line(f"  ... and {len(bundles) - 5} more patches cataloged in remediation.md"))
-        print(border_bottom())
-        print()
-
-        print(border_top("Next Governed Action", border_color=GREEN, double=True))
-        print(box_line(f"Review & Apply:   {BOLD}{WHITE}npx torusguard apply{RESET}  (CLI Human Gate)", border="║", border_color=GREEN))
-        print(box_line(f"Auto-Apply Flag:  {CYAN}npx torusguard apply --yes{RESET} (Automated mode)", border="║", border_color=GREEN))
-        print(box_line(f"AI IDE Chat:      Run {CYAN}/torusguard-apply{RESET} in your AI chat", border="║", border_color=GREEN))
-        print(border_bottom(border_color=GREEN, double=True))
-        print()
-    else:
-        print(f"  {YELLOW}ℹ No automatic patch templates matched current findings.{RESET}")
-        print(f"  {GRAY}Findings require architectural refactoring or AI-assisted guidance.{RESET}\n")
-        print(border_top("Recommended Next Steps"))
-        print(box_line("1. In AI Chat: Run /torusguard-harden to synthesize custom fixes"))
-        print(box_line("2. View visual posture report: npx torusguard report --html"))
-        print(box_line(f"3. Inspect findings: .torusguard/runs/{run_folder.name}/findings.md"))
-        print(border_bottom())
-        print()
-
-    return {
-        "status": "success",
-        "run_folder": str(run_folder),
-        "bundles_count": len(bundles),
-        "bundles": bundles
-    }
-
-
-def main():
-    parser = argparse.ArgumentParser(description="TorusGuard Autonomous Governed Remediation Engine")
-    parser.add_argument("path", nargs="?", default=".", help="Target project root directory")
-    parser.add_argument("--run", "-r", help="Explicit run ID to harden (default: latest audit run)")
-    parser.add_argument("--json", action="store_true", help="Output raw JSON")
-    args = parser.parse_args()
-
-    target = Path(args.path).resolve()
-    result = execute_harden(target, run_id_arg=args.run)
-
-    if args.json:
-        print(json.dumps(result, indent=2))
-
-
-if __name__ == "__main__":
-    main()
+#!/usr/bin/env python3
+"""
+TorusGuard Autonomous Remediation Engine (v1.3.5)
+Evaluates audit findings against canonical patch templates, enforces Ponytail bounds
+(<= 35 additions, <= 25 deletions), packages structured candidate remediation bundles,
+and renders unified patch diffs with standardized 75-column terminal UI.
+
+Pure Python 3.10+ standard library (zero external dependencies).
+"""
+
+import sys
+import os
+import re
+import json
+import difflib
+import hashlib
+import argparse
+import datetime
+from pathlib import Path
+from typing import Dict, List, Any, Optional
+
+# Ensure scripts dir is in sys.path for term_ui import
+scripts_dir = Path(__file__).resolve().parent
+if str(scripts_dir) not in sys.path:
+    sys.path.insert(0, str(scripts_dir))
+
+try:
+    import term_ui
+except ImportError:
+    term_ui = None
+
+# Windows console UTF-8 support
+if sys.stdout and hasattr(sys.stdout, "reconfigure"):
+    try:
+        getattr(sys.stdout, "reconfigure")(encoding="utf-8", errors="replace")
+    except Exception:
+        pass
+
+# ─── Timezone Configuration: Indian Standard Time (IST, UTC+05:30) ───────────
+IST = datetime.timezone(datetime.timedelta(hours=5, minutes=30), name="IST")
+
+def get_ist_now() -> datetime.datetime:
+    return datetime.datetime.now(IST)
+
+# ─── Fallback Formatters (if term_ui missing) ──────────────────────────────────
+BOLD = "\033[1m"
+DIM = "\033[2m"
+RESET = "\033[0m"
+CYAN = "\033[36m"
+GREEN = "\033[32m"
+YELLOW = "\033[33m"
+WHITE = "\033[97m"
+GRAY = "\033[90m"
+RED = "\033[31m"
+
+def box_line(content: str, width: int = 67, border: str = "│", border_color: str = CYAN) -> str:
+    if term_ui:
+        return term_ui.format_box_line(content, width=width, border=border, border_color=border_color)
+    return f"  {border_color}{border}{RESET}  {content}"
+
+def box_header(title: str, subtitle: str = "", version: str = "v1.3.5", border_color: str = CYAN) -> str:
+    if term_ui:
+        return term_ui.card_header(title, subtitle, version, border_color)
+    return f"=== {title} ({version}) ==="
+
+def border_top(title: str = "", border_color: str = CYAN, double: bool = False) -> str:
+    if term_ui:
+        return term_ui.card_border_top(title, border_color=border_color, double=double)
+    return "┌" + "─" * 71 + "┐"
+
+def border_bottom(border_color: str = CYAN, double: bool = False) -> str:
+    if term_ui:
+        return term_ui.card_border_bottom(border_color=border_color, double=double)
+    return "└" + "─" * 71 + "┘"
+
+
+def find_latest_audit_run(runs_dir: Path) -> Optional[Path]:
+    """Locate the most recent audit run folder containing findings.json."""
+    if not runs_dir.is_dir():
+        return None
+    runs = sorted(runs_dir.glob("run-*-audit"), key=lambda p: p.stat().st_mtime, reverse=True)
+    for r in runs:
+        if (r / "findings.json").is_file():
+            return r
+    return None
+
+
+def generate_patch_for_finding(finding: Dict[str, Any], target_root: Path) -> Optional[Dict[str, Any]]:
+    """
+    Formulate a minimal, surgical fix for a finding adhering to Ponytail bounds
+    (<= 35 additions, <= 25 deletions).
+    """
+    rule_id = finding.get("rule_id", "")
+    file_path = finding.get("file_path", "")
+    line_number = finding.get("line_number", 1)
+    description = finding.get("description", "")
+
+    full_target = target_root / file_path
+    if not full_target.is_file():
+        return None
+
+    try:
+        content = full_target.read_text(encoding="utf-8", errors="replace")
+    except Exception:
+        return None
+
+    lines = content.splitlines()
+    if not (1 <= line_number <= len(lines)):
+        return None
+
+    target_line = lines[line_number - 1]
+    indent = re.match(r'^\s*', target_line).group(0)  # type: ignore
+    new_lines = list(lines)
+    applied_rule = False
+    fix_explanation = ""
+
+    # Strategy 1: Hardcoded Secrets (TG-SEC-001 / TG-SEC-002)
+    if rule_id in ("TG-SEC-001", "TG-SEC-002"):
+        if file_path.endswith((".js", ".jsx", ".ts", ".tsx", ".mjs")):
+            if re.search(r'(jwt_secret|jwtSecret|JWT_SECRET)', target_line, re.IGNORECASE):
+                var_match = re.search(r'(const|let|var)\s+([A-Za-z0-9_]+)\s*=', target_line)
+                var_name = var_match.group(2) if var_match else "JWT_SECRET"
+                replacement = f'{indent}const {var_name} = process.env.JWT_SECRET || "";'
+                new_lines[line_number - 1] = replacement
+                applied_rule = True
+                fix_explanation = f"Migrated hardcoded JWT secret to environment variable process.env.JWT_SECRET"
+            elif re.search(r'(:=|=)\s*["\'][A-Za-z0-9_\-]{8,}["\']', target_line):
+                var_match = re.search(r'(const|let|var)?\s*([A-Za-z0-9_]+)\s*[:=]', target_line)
+                var_name = var_match.group(2) if var_match else "API_KEY"
+                env_key = re.sub(r'(?<!^)(?=[A-Z])', '_', var_name).upper()
+                replacement = f'{indent}const {var_name} = process.env.{env_key} || "";'
+                new_lines[line_number - 1] = replacement
+                applied_rule = True
+                fix_explanation = f"Replaced hardcoded credential with process.env.{env_key}"
+            elif "password:" in target_line or "password :" in target_line:
+                fixed_line = re.sub(r'password:\s*["\'][^"\']+["\']', 'password: process.env.DEMO_USER_PASSWORD || "REDACTED_PASSWORD"', target_line)
+                if fixed_line != target_line:
+                    new_lines[line_number - 1] = fixed_line
+                    applied_rule = True
+                    fix_explanation = "Replaced hardcoded password with environment variable process.env.DEMO_USER_PASSWORD"
+        elif file_path.endswith(".py"):
+            if re.search(r'(:=|=)\s*["\'][A-Za-z0-9_\-]{8,}["\']', target_line):
+                var_match = re.search(r'([A-Za-z0-9_]+)\s*=', target_line)
+                var_name = var_match.group(1) if var_match else "SECRET_KEY"
+                env_key = var_name.upper()
+                replacement = f'{indent}{var_name} = os.environ.get("{env_key}", "")'
+                new_lines[line_number - 1] = replacement
+                applied_rule = True
+                fix_explanation = f"Replaced hardcoded credential with os.environ.get('{env_key}')"
+
+    # Strategy 2: Dangerous React HTML rendering & DOM innerHTML (TG-INPUT-003)
+    elif rule_id == "TG-INPUT-003":
+        if "dangerouslySetInnerHTML" in target_line:
+            var_match = re.search(r'__html\s*:\s*([A-Za-z0-9_$.]+)', target_line)
+            var_name = var_match.group(1) if var_match else "content"
+            replacement = f'{indent}<div>{{{var_name}}}</div>'
+            new_lines[line_number - 1] = replacement
+            applied_rule = True
+            fix_explanation = "Replaced raw dangerouslySetInnerHTML injection with safe React text interpolation"
+        elif ".innerHTML" in target_line:
+            if re.search(r'\.innerHTML\s*=\s*["\']["\']', target_line):
+                fixed_line = re.sub(r'([a-zA-Z0-9_$]+)\.innerHTML\s*=\s*["\']["\']', r'\1.textContent = ""', target_line)
+                if fixed_line != target_line:
+                    new_lines[line_number - 1] = fixed_line
+                    applied_rule = True
+                    fix_explanation = "Replaced unsafe innerHTML DOM reset with safe textContent assignment"
+            elif re.search(r'([a-zA-Z0-9_$]+)\.innerHTML\s*=\s*([a-zA-Z0-9_$.]+)\s*;?$', target_line):
+                fixed_line = re.sub(r'([a-zA-Z0-9_$]+)\.innerHTML\s*=\s*([a-zA-Z0-9_$.]+)', r'\1.textContent = \2', target_line)
+                if fixed_line != target_line:
+                    new_lines[line_number - 1] = fixed_line
+                    applied_rule = True
+                    fix_explanation = "Replaced raw DOM innerHTML assignment with XSS-safe textContent"
+
+    # Strategy 3: Raw SQL Concatenation (TG-INPUT-002)
+    elif rule_id == "TG-INPUT-002":
+        if file_path.endswith(".py") and ("f\"SELECT" in target_line or "f'SELECT" in target_line):
+            fixed_line = re.sub(r'f(["\']SELECT.*?WHERE.*?)(\{[^}]+\})', r'\1%s', target_line)
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Replaced unsafe Python SQL f-string interpolation with parameterized SQL query placeholder"
+        elif file_path.endswith((".js", ".jsx", ".ts", ".tsx", ".mjs")):
+            # Template string SQL injection: `SELECT ... ${var}` -> parameterized placeholder
+            if "`SELECT" in target_line or "`select" in target_line:
+                fixed_line = re.sub(r'\$\{[^}]+\}', '?', target_line)
+                if fixed_line != target_line:
+                    new_lines[line_number - 1] = fixed_line
+                    applied_rule = True
+                    fix_explanation = "Replaced template literal SQL string interpolation with parameterized SQL placeholder (?)"
+
+    # Strategy 4: Multi-Tenant Query Scoping (TG-DB-004)
+    elif rule_id == "TG-DB-004":
+        if ".objects.get(" in target_line and "tenant" not in target_line.lower() and "org" not in target_line.lower():
+            fixed_line = target_line.replace(".objects.get(", ".objects.get(organization_id=request.user.organization_id, ")
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Added explicit organization_id tenant boundary filter to ORM query"
+        elif "findUnique(" in target_line and "tenant" not in target_line.lower():
+            fixed_line = target_line.replace("findUnique({ where: {", "findFirst({ where: { tenantId: req.tenantId,")
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Enforced tenantId multi-tenant isolation in Prisma lookup"
+
+    # Strategy 5: Permissive CORS with Credentials (TG-PLATFORM-001)
+    elif rule_id == "TG-PLATFORM-001":
+        if "CORS_ALLOW_ALL_ORIGINS = True" in target_line:
+            replacement = f'{indent}CORS_ALLOWED_ORIGINS = [os.environ.get("ALLOWED_ORIGIN", "https://app.example.com")]'
+            new_lines[line_number - 1] = replacement
+            applied_rule = True
+            fix_explanation = "Replaced wildcard CORS configuration with strict origin whitelist environment variable"
+        elif "cors(" in target_line and ("origin: '*'" in target_line or 'origin: "*"' in target_line):
+            fixed_line = target_line.replace("origin: '*'", "origin: process.env.ALLOWED_ORIGIN || 'http://localhost:5173'")
+            fixed_line = fixed_line.replace('origin: "*"', "origin: process.env.ALLOWED_ORIGIN || 'http://localhost:5173'")
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Constrained wildcard CORS origin to process.env.ALLOWED_ORIGIN with secure fallback"
+
+    # Strategy 6: Cookie Missing Security Flags (TG-AUTH-004)
+    elif rule_id == "TG-AUTH-004":
+        if "res.cookie(" in target_line and "httpOnly" not in target_line:
+            # res.cookie('token', token) -> res.cookie('token', token, { httpOnly: true, secure: process.env.NODE_ENV === 'production', sameSite: 'lax' })
+            fixed_line = re.sub(
+                r'res\.cookie\(\s*([a-zA-Z0-9_\'\"]+)\s*,\s*([a-zA-Z0-9_]+)\s*\)',
+                r"res.cookie(\1, \2, { httpOnly: true, secure: process.env.NODE_ENV === 'production', sameSite: 'lax' })",
+                target_line
+            )
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Injected mandatory security flags (httpOnly, secure, sameSite) to cookie definition"
+
+    # Strategy 7: Path Traversal (TG-INPUT-006)
+    elif rule_id == "TG-INPUT-006":
+        if "path.join(" in target_line and "path.basename" not in target_line:
+            fixed_line = re.sub(r'path\.join\((.*?),\s*([a-zA-Z0-9_$.]+)\)', r'path.join(\1, path.basename(\2))', target_line)
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Wrapped dynamic path segment in path.basename to neutralize path traversal sequences"
+        elif "os.path.join(" in target_line and "os.path.basename" not in target_line:
+            fixed_line = re.sub(r'os\.path\.join\((.*?),\s*([a-zA-Z0-9_$.]+)\)', r'os.path.join(\1, os.path.basename(\2))', target_line)
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Wrapped dynamic path segment in os.path.basename to neutralize path traversal sequences"
+
+    # Strategy 8: TLS Verification Bypass (TG-DIFF-001)
+    elif rule_id == "TG-DIFF-001":
+        if "verify=False" in target_line:
+            fixed_line = target_line.replace("verify=False", "verify=True")
+            new_lines[line_number - 1] = fixed_line
+            applied_rule = True
+            fix_explanation = "Restored mandatory TLS certificate verification (verify=True)"
+        elif "rejectUnauthorized: false" in target_line:
+            fixed_line = target_line.replace("rejectUnauthorized: false", "rejectUnauthorized: true")
+            new_lines[line_number - 1] = fixed_line
+            applied_rule = True
+            fix_explanation = "Restored TLS certificate authority validation (rejectUnauthorized: true)"
+        elif "InsecureSkipVerify: true" in target_line:
+            fixed_line = target_line.replace("InsecureSkipVerify: true", "InsecureSkipVerify: false")
+            new_lines[line_number - 1] = fixed_line
+            applied_rule = True
+            fix_explanation = "Restored mandatory TLS certificate validation (InsecureSkipVerify: false)"
+
+
+    # Strategy 9: Hardcoded Database / Service Credential in Config (TG-SEC-003)
+    elif rule_id == "TG-SEC-003":
+        if "=" in target_line:
+            var_match = re.match(r'^\s*([A-Za-z0-9_]+)\s*=', target_line)
+            if var_match:
+                vname = var_match.group(1)
+                fixed_line = f'{indent}{vname} = os.environ.get("{vname}", "")' if file_path.endswith('.py') else f'{indent}const {vname} = process.env.{vname} || "";'
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = f"Extracted sensitive config variable {vname} to environment variable"
+
+    # Strategy 10: Weak Password Hashing (TG-AUTH-001)
+    elif rule_id == "TG-AUTH-001":
+        if "hashlib.md5" in target_line:
+            fixed_line = target_line.replace("hashlib.md5(", "hashlib.sha256(")
+            new_lines[line_number - 1] = fixed_line
+            applied_rule = True
+            fix_explanation = "Replaced broken MD5 hashing with SHA-256 (or bcrypt/argon2)"
+        elif "createHash('md5')" in target_line:
+            fixed_line = target_line.replace("createHash('md5')", "createHash('sha256')")
+            new_lines[line_number - 1] = fixed_line
+            applied_rule = True
+            fix_explanation = "Replaced broken MD5 hashing with SHA-256"
+
+    # Strategy 11: JWT Missing Algorithm Restriction (TG-AUTH-003)
+    elif rule_id == "TG-AUTH-003":
+        if "jwt.verify(" in target_line and "algorithms" not in target_line:
+            fixed_line = re.sub(r'jwt\.verify\(\s*([^,\)]+)\s*,\s*([^,\)]+)\s*\)', r"jwt.verify(\1, \2, { algorithms: ['HS256'] })", target_line)
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Restricted JWT verification to explicit algorithm list ['HS256']"
+
+    # Strategy 12: Untrusted Role/Tenant Header Injection (TG-AUTH-008)
+    elif rule_id == "TG-AUTH-008":
+        if "req.headers[" in target_line or "request.headers.get(" in target_line:
+            if file_path.endswith((".js", ".ts")):
+                fixed_line = f'{indent}const role = req.user ? req.user.role : "user"; // Derived securely from authenticated session'
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Replaced unverified client HTTP header with verified session-derived identity"
+            elif file_path.endswith(".py"):
+                fixed_line = f'{indent}role = getattr(request.user, "role", "user") # Derived securely from authenticated session'
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Replaced unverified client HTTP header with verified session-derived identity"
+
+    # Strategy 13: Rate Limiting Injection (TG-RATE-001)
+    elif rule_id == "TG-RATE-001":
+        if re.search(r'\.(?:post|all)\s*\(\s*[\'"][^\'"]+[\'"]\s*,\s*(?:async\s*)?\(', target_line):
+            fixed_line = re.sub(
+                r'(\.(?:post|all)\s*\(\s*[\'"][^\'"]+[\'"]\s*),\s*((?:async\s*)?\()',
+                r'\1, authLimiter, \2',
+                target_line
+            )
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Injected authLimiter rate-limiting middleware to protect authentication endpoint from brute force"
+
+    # Strategy 14: Prompt Injection System Context Isolation (TG-AGENT-001)
+    elif rule_id == "TG-AGENT-001":
+        if 'role": "system"' in target_line or "role': 'system'" in target_line:
+            fixed_line = target_line.replace('"role": "system"', '"role": "user"')
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Demoted untrusted user input from system prompt into role: 'user' message container"
+
+    # Strategy 15: Outbound URL Fetch Validation (TG-SSRF-001)
+    elif rule_id == "TG-SSRF-001":
+        if "fetch(" in target_line and "validateUrl(" not in target_line:
+            fixed_line = re.sub(r'fetch\(\s*([^,\)]+)', r'fetch(validateSafeUrl(\1)', target_line)
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Wrapped outbound URL in validateSafeUrl domain whitelist guard"
+        elif "requests.get(" in target_line and "validate_url(" not in target_line:
+            fixed_line = re.sub(r'requests\.get\(\s*([^,\)]+)', r'requests.get(validate_safe_url(\1)', target_line)
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Wrapped outbound URL in validate_safe_url domain whitelist guard"
+
+    # Strategy 16: Outbound Request Timeout (TG-SSRF-004)
+    elif rule_id == "TG-SSRF-004":
+        if "requests.get(" in target_line and "timeout=" not in target_line:
+            fixed_line = target_line.replace(")", ", timeout=10)")
+            new_lines[line_number - 1] = fixed_line
+            applied_rule = True
+            fix_explanation = "Injected explicit 10-second request timeout to prevent resource exhaustion"
+
+    # Strategy 17: Production Debug Mode Disablement (TG-PLATFORM-003)
+    elif rule_id == "TG-PLATFORM-003":
+        if "DEBUG = True" in target_line:
+            fixed_line = f'{indent}DEBUG = os.environ.get("DJANGO_DEBUG", "False").lower() in ("true", "1")'
+            new_lines[line_number - 1] = fixed_line
+            applied_rule = True
+            fix_explanation = "Guard DEBUG mode behind DJANGO_DEBUG environment variable"
+
+    # Strategy 18: Production Introspection Disablement (TG-GQL-004)
+    elif rule_id == "TG-GQL-004":
+        if "introspection: true" in target_line or "introspection: True" in target_line:
+            fixed_line = re.sub(r'introspection\s*:\s*(?:true|True)', 'introspection: process.env.NODE_ENV !== "production"', target_line)
+            new_lines[line_number - 1] = fixed_line
+    # Strategy 19: Privileged Database Credential in Frontend (TG-DB-002)
+    elif rule_id == "TG-DB-002":
+        if "SUPABASE_SERVICE_ROLE_KEY" in target_line:
+            var_match = re.search(r'(export\s+const|const)\s+([A-Za-z0-9_]+)\s*=', target_line)
+            var_name = var_match.group(2) if var_match else "SUPABASE_SERVICE_ROLE_KEY"
+            replacement = f'{indent}export const {var_name} = process.env.{var_name} || "";'
+            new_lines[line_number - 1] = replacement
+            applied_rule = True
+            fix_explanation = f"Replaced hardcoded privileged database credential with process.env.{var_name}"
+
+    # Strategy 20: Public Source Maps in Production (TG-CLIENT-001)
+    elif rule_id == "TG-CLIENT-001":
+        if "sourcemap: true" in target_line:
+            fixed_line = target_line.replace("sourcemap: true", "sourcemap: false")
+            new_lines[line_number - 1] = fixed_line
+            applied_rule = True
+            fix_explanation = "Disabled public production source map emission"
+
+    # Strategy 21: Sensitive Information in Logs (TG-SEC-004)
+    elif rule_id == "TG-SEC-004":
+        if "console.log" in target_line:
+            fixed_line = f"{indent}// Security: Sensitive credential log removed per TG-SEC-004"
+            new_lines[line_number - 1] = fixed_line
+            applied_rule = True
+            fix_explanation = "Redacted console logging of sensitive token or credential"
+
+    # Strategy 22: Unlimited Public Write Endpoint (TG-RATE-002)
+    elif rule_id == "TG-RATE-002":
+        if re.search(r'\.(?:post|put|patch)\s*\(\s*[\'"][^\'"]+[\'"]\s*,\s*(?:async\s*)?\(', target_line):
+            fixed_line = re.sub(
+                r'(\.(?:post|put|patch)\s*\(\s*[\'"][^\'"]+[\'"]\s*),\s*((?:async\s*)?\()',
+                r'\1, writeLimiter, \2',
+                target_line
+            )
+            if fixed_line != target_line:
+                new_lines[line_number - 1] = fixed_line
+                applied_rule = True
+                fix_explanation = "Injected writeLimiter rate-limiting middleware to public write endpoint"
+
+    # Strategy 23: Missing Security Headers (TG-PLATFORM-002)
+    elif rule_id == "TG-PLATFORM-002":
+        if "const app = express();" in target_line:
+            replacement = f'{indent}const helmet = require("helmet");\n{indent}const app = express();\n{indent}app.use(helmet());'
+            new_lines[line_number - 1] = replacement
+            applied_rule = True
+            fix_explanation = "Injected helmet security headers middleware into Express application"
+
+    # Strategy 24: Missing Request Body Validation (TG-INPUT-001)
+    elif rule_id == "TG-INPUT-001":
+        if "req.body" in target_line and ("const {" in target_line or "let {" in target_line):
+            replacement = f'{indent}if (!req.body || typeof req.body !== "object") return res.status(400).json({{ error: "Invalid payload" }});\n{target_line}'
+            new_lines[line_number - 1] = replacement
+            applied_rule = True
+            fix_explanation = "Added defensive request body payload validation before destructuring"
+
+    if not applied_rule:
+        return None
+
+    # Calculate unified diff
+    orig_lines = [l + "\n" for l in lines]
+    patched_lines = [l + "\n" for l in new_lines]
+    diff = "".join(difflib.unified_diff(
+        orig_lines, patched_lines,
+        fromfile=f"a/{file_path}", tofile=f"b/{file_path}",
+        n=2
+    ))
+
+    additions = sum(1 for l in diff.splitlines() if l.startswith("+") and not l.startswith("+++"))
+    deletions = sum(1 for l in diff.splitlines() if l.startswith("-") and not l.startswith("---"))
+
+    # Ponytail bounds validation: <= 35 additions, <= 25 deletions
+    if additions > 35 or deletions > 25:
+        return None
+
+    bundle_id = f"bnd-{rule_id.lower()}-{line_number}-{hashlib.sha256(diff.encode('utf-8')).hexdigest()[:6]}"
+
+    return {
+        "bundle_id": bundle_id,
+        "finding_id": finding.get("finding_id", f"{rule_id}-{line_number}"),
+        "rule_id": rule_id,
+        "title": finding.get("title", rule_id),
+        "target_file": file_path,
+        "line_number": line_number,
+        "target_line": target_line,
+        "replacement_line": new_lines[line_number - 1],
+        "what_is_wrong": description,
+        "why_it_matters": "Security vulnerability violating TorusGuard strict production safety invariant.",
+        "what_should_change": fix_explanation,
+        "proposed_diff": diff,
+        "additions": additions,
+        "deletions": deletions,
+        "patched_content": "\n".join(new_lines)
+    }
+
+
+def execute_harden(target_root: Path, run_id_arg: Optional[str] = None) -> Dict[str, Any]:
+    """Execute autonomous remediation formulation and package candidate bundles."""
+    runs_dir = target_root / ".torusguard" / "runs"
+    if run_id_arg:
+        run_folder = runs_dir / run_id_arg
+    else:
+        run_folder = find_latest_audit_run(runs_dir)
+
+    if not run_folder or not run_folder.is_dir():
+        print(f"\n  {RED}✖ No audit runs found to harden.{RESET}")
+        print(f"  Run {CYAN}npx torusguard audit{RESET} first to discover actionable findings.\n")
+        sys.exit(1)
+
+    findings_file = run_folder / "findings.json"
+    if not findings_file.is_file():
+        print(f"\n  {RED}✖ Missing findings.json in run folder:{RESET} {run_folder}\n")
+        sys.exit(1)
+
+    try:
+        findings = json.loads(findings_file.read_text(encoding="utf-8"))
+    except Exception as e:
+        print(f"\n  {RED}✖ Failed to parse findings.json:{RESET} {e}\n")
+        sys.exit(1)
+
+    bundles_dir = run_folder / "bundles"
+    bundles_dir.mkdir(parents=True, exist_ok=True)
+
+    bundles = []
+    seen_file_lines = set()
+    for f in findings:
+        target_key = (f.get("file_path"), f.get("line_number"))
+        if target_key in seen_file_lines:
+            continue
+        candidate = generate_patch_for_finding(f, target_root)
+        if candidate:
+            seen_file_lines.add(target_key)
+            bundles.append(candidate)
+            b_dir = bundles_dir / candidate["bundle_id"]
+            b_dir.mkdir(parents=True, exist_ok=True)
+
+            # 1. patch.diff
+            (b_dir / "patch.diff").write_text(candidate["proposed_diff"], encoding="utf-8")
+
+            # 2. minimal_patch_plan.md
+            plan_content = [
+                f"# Remediation Plan: {candidate['bundle_id']}",
+                f"- **Rule ID:** `{candidate['rule_id']}`",
+                f"- **Target File:** `{candidate['target_file']}:{candidate['line_number']}`",
+                f"- **Ponytail Churn:** `+{candidate['additions']} / -{candidate['deletions']}` (Compliant <=35/<=25)",
+                f"\n## Proposed Change\n{candidate['what_should_change']}\n",
+                "## Unified Diff Preview\n```diff",
+                candidate["proposed_diff"],
+                "```"
+            ]
+            (b_dir / "minimal_patch_plan.md").write_text("\n".join(plan_content), encoding="utf-8")
+
+            # 3. metadata.json
+            with open(b_dir / "metadata.json", "w", encoding="utf-8") as mf:
+                json.dump(candidate, mf, indent=2)
+
+    # Render run-level remediation.md
+    remediation_lines = [
+        f"# TorusGuard Governed Remediation Catalog",
+        f"**Run ID:** `{run_folder.name}`  ",
+        f"**Generated:** `{get_ist_now().strftime('%Y-%m-%d %H:%M:%S IST')}`  ",
+        f"**Candidate Bundles Formulated:** `{len(bundles)}`  \n",
+        "---",
+        "## Formulated Candidate Patches\n"
+    ]
+
+    for b in bundles:
+        remediation_lines.append(f"### [{b['rule_id']}] {b['title']}")
+        remediation_lines.append(f"- **Target:** `{b['target_file']}:{b['line_number']}` | **Bundle ID:** `{b['bundle_id']}`")
+        remediation_lines.append(f"- **Strategy:** {b['what_should_change']}")
+        remediation_lines.append(f"- **Ponytail Churn:** `+{b['additions']} / -{b['deletions']}`")
+        remediation_lines.append("```diff")
+        remediation_lines.append(b["proposed_diff"].strip())
+        remediation_lines.append("```\n")
+
+    (run_folder / "remediation.md").write_text("\n".join(remediation_lines), encoding="utf-8")
+    try:
+        import report_sync
+        report_sync.record_harden_bundles(target_root, bundles, run_folder.name)
+    except Exception:
+        pass
+
+    # Update manifest
+    manifest_file = run_folder / "manifest.json"
+    if manifest_file.is_file():
+        try:
+            mdata = json.loads(manifest_file.read_text(encoding="utf-8"))
+            mdata["remediation_bundles_count"] = len(bundles)
+            manifest_file.write_text(json.dumps(mdata, indent=2), encoding="utf-8")
+        except Exception:
+            pass
+
+    # Print 75-column Terminal Cards
+    print()
+    print(box_header("🛡️  TORUSGUARD GOVERNED REMEDIATION ENGINE", "Autonomous Minimal Patch Formulation & Ponytail Packaging", "v1.3.5"))
+    print()
+
+    print(border_top("Remediation Scope"))
+    print(box_line(f"Active Run:     {WHITE}{run_folder.name}{RESET}"))
+    print(box_line(f"Target Scope:   {WHITE}{target_root}{RESET}"))
+    print(box_line(f"Total Findings: {len(findings)} evaluated | {GREEN}{len(bundles)} surgical patches formulated{RESET}"))
+    print(box_line(f"Ponytail Guard: {GREEN}All patches strictly <= 35 additions, <= 25 deletions{RESET}"))
+    print(border_bottom())
+    print()
+
+    if bundles:
+        print(border_top(f"Formulated Candidate Patches ({len(bundles)})"))
+        for b in bundles[:5]:
+            print(box_line(f"{YELLOW}[{b['rule_id']}]{RESET} {WHITE}{b['target_file']}:{b['line_number']}{RESET} (+{b['additions']}/-{b['deletions']})"))
+            print(box_line(f"  └─ {DIM}{b['what_should_change']}{RESET}"))
+        if len(bundles) > 5:
+            print(box_line(f"  ... and {len(bundles) - 5} more patches cataloged in remediation.md"))
+        print(border_bottom())
+        print()
+
+        print(border_top("Next Governed Action", border_color=GREEN, double=True))
+        print(box_line(f"Living Report:    {CYAN}security_report.md (patch previews attached){RESET}", border="║", border_color=GREEN))
+        print(box_line(f"Review & Apply:   {BOLD}{WHITE}npx torusguard apply{RESET}  (CLI Human Gate)", border="║", border_color=GREEN))
+        print(box_line(f"Auto-Apply Flag:  {CYAN}npx torusguard apply --yes{RESET} (Automated mode)", border="║", border_color=GREEN))
+        print(box_line(f"AI IDE Chat:      Run {CYAN}/torusguard-apply{RESET} in your AI chat", border="║", border_color=GREEN))
+        print(border_bottom(border_color=GREEN, double=True))
+        print()
+    else:
+        print(f"  {YELLOW}ℹ No automatic patch templates matched current findings.{RESET}")
+        print(f"  {GRAY}Findings require architectural refactoring or AI-assisted guidance.{RESET}\n")
+        print(border_top("Recommended Next Steps"))
+        print(box_line("1. In AI Chat: Run /torusguard-harden to synthesize custom fixes"))
+        print(box_line("2. View visual posture report: npx torusguard report --html"))
+        print(box_line(f"3. Inspect findings: .torusguard/runs/{run_folder.name}/findings.md"))
+        print(border_bottom())
+        print()
+
+    return {
+        "status": "success",
+        "run_folder": str(run_folder),
+        "bundles_count": len(bundles),
+        "bundles": bundles
+    }
+
+
+def main():
+    parser = argparse.ArgumentParser(description="TorusGuard Autonomous Governed Remediation Engine")
+    parser.add_argument("path", nargs="?", default=".", help="Target project root directory")
+    parser.add_argument("--run", "-r", help="Explicit run ID to harden (default: latest audit run)")
+    parser.add_argument("--json", action="store_true", help="Output raw JSON")
+    args = parser.parse_args()
+
+    target = Path(args.path).resolve()
+    result = execute_harden(target, run_id_arg=args.run)
+
+    if args.json:
+        print(json.dumps(result, indent=2))
+
+
+if __name__ == "__main__":
+    main()