torusguard 2.1.0 → 2.1.2

package/skills/torusguard/payload/core/__init__.py

@@ -0,0 +1,146 @@
+"""
+TorusGuard Core Architecture & Security Workflow Engine (v0.7.0)
+
+Tiered architectural layers:
+- Tier 1: Canonical Data Models, Lifecycle State Machine & Reporting (v0.5 base)
+- Tier 2: Governed Remediation, Clustering, Patch Policy & SARIF Export (v0.6 engine)
+- Tier 3: Authorized Runtime Validation, Exploitability Confirmation & Multi-Agent Roles (v0.7 engine)
+"""
+
+# ==============================================================================
+# Tier 1: Canonical Data Models, Lifecycle & Provenance (v0.5 base)
+# ==============================================================================
+from .models import (
+    Finding,
+    Evidence,
+    Remediation,
+    FrameworkPattern,
+    AffectedComponent,
+    ReproductionMethod,
+    RetestRecord,
+    NotesRecord,
+    FindingTimestamps,
+    ProvenanceChain,
+    ConfidenceScore,
+    ConfidenceFactors,
+    ConfidenceBand,
+    SeverityLevel,
+    SeverityInfo,
+    RemediationPriority,
+    FindingStatus,
+    LifecycleStage,
+    TaxonomyCategory,
+    EvidenceType,
+    AuditReport,
+    mask_sensitive_data,
+)
+from .lifecycle import FindingLifecycleManager, LifecycleTransitionError
+from .formatter import ReportFormatter
+from .run_folder import RunFolder
+
+# ==============================================================================
+# Tier 2: Governed Remediation, Clustering, Patch Policy & SARIF (v0.6 engine)
+# ==============================================================================
+from .identity import IdentityEngine, FindingFingerprint
+from .clustering import ClusteringEngine, RootCauseCluster
+from .bundle import BundleManager, RemediationBundle
+from .governance import PatchGovernor, PatchPolicyDecision, HIGH_RISK_KEYWORDS, SENSITIVE_CATEGORIES
+from .rechecker import TargetedRechecker, TargetedRecheckResult, RecheckOutcome
+from .run_manager import RunManager
+from .sarif import SarifExporter
+from .v6_reporter import V6Reporter
+from .v6_workflow import V6Workflow
+from .stack_profiler import StackProfiler, StackProfile
+
+# ==============================================================================
+# Tier 3: Authorized Runtime Validation & Exploitability Confirmation (v0.7 engine)
+# ==============================================================================
+from .authorization import AuthorizationManager, AuthorizationRecord, TargetScope, AuthorizationError
+from .safety_gate import SafetyGate, SafetyDecision, SafetyReviewLevel
+from .runtime_evidence import EvidenceCollector, RuntimeEvidenceItem, RedactionEngine
+from .runtime_validator import WebValidator, SessionState
+from .exploit_checker import ExploitChecker, ExploitCheckResult, ExploitabilityStatus
+from .browser_verifier import BrowserVerifier, BrowserAction
+from .agent_roles import RoleOrchestrator, RoleHandoff, AgentRole
+from .replay_trace import ReplayManager, ReplayTrace, ReplayStep
+from .v070_reporter import V070Reporter
+from .v070_workflow import V070Workflow
+
+__version__ = "0.7.0"
+
+__all__ = [
+    # Tier 1: Canonical Models & Lifecycle
+    "Finding",
+    "Evidence",
+    "Remediation",
+    "FrameworkPattern",
+    "AffectedComponent",
+    "ReproductionMethod",
+    "RetestRecord",
+    "NotesRecord",
+    "FindingTimestamps",
+    "ProvenanceChain",
+    "ConfidenceScore",
+    "ConfidenceFactors",
+    "ConfidenceBand",
+    "SeverityLevel",
+    "SeverityInfo",
+    "RemediationPriority",
+    "FindingStatus",
+    "LifecycleStage",
+    "TaxonomyCategory",
+    "EvidenceType",
+    "AuditReport",
+    "mask_sensitive_data",
+    "FindingLifecycleManager",
+    "LifecycleTransitionError",
+    "ReportFormatter",
+    "RunFolder",
+    # Tier 2: Governed Remediation & v0.6 Engine
+    "IdentityEngine",
+    "FindingFingerprint",
+    "ClusteringEngine",
+    "RootCauseCluster",
+    "BundleManager",
+    "RemediationBundle",
+    "PatchGovernor",
+    "PatchPolicyDecision",
+    "HIGH_RISK_KEYWORDS",
+    "SENSITIVE_CATEGORIES",
+    "TargetedRechecker",
+    "TargetedRecheckResult",
+    "RecheckOutcome",
+    "RunManager",
+    "SarifExporter",
+    "V6Reporter",
+    "V6Workflow",
+    "StackProfiler",
+    "StackProfile",
+    # Tier 3: Authorized Runtime Validation & v0.7 Engine
+    "AuthorizationManager",
+    "AuthorizationRecord",
+    "TargetScope",
+    "AuthorizationError",
+    "SafetyGate",
+    "SafetyDecision",
+    "SafetyReviewLevel",
+    "EvidenceCollector",
+    "RuntimeEvidenceItem",
+    "RedactionEngine",
+    "WebValidator",
+    "SessionState",
+    "ExploitChecker",
+    "ExploitCheckResult",
+    "ExploitabilityStatus",
+    "BrowserVerifier",
+    "BrowserAction",
+    "RoleOrchestrator",
+    "RoleHandoff",
+    "AgentRole",
+    "ReplayManager",
+    "ReplayTrace",
+    "ReplayStep",
+    "V070Reporter",
+    "V070Workflow",
+]
+

package/skills/torusguard/payload/core/agent_roles.py

@@ -0,0 +1,104 @@
+"""
+TorusGuard v0.7.0 Role-Based Multi-Agent Workflow Engine
+Coordinates explicit handoff contracts between Profiler, Validator, Remediator,
+and Reviewer agent roles, ensuring strict authority separation and full auditability.
+"""
+
+import json
+from enum import Enum
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Dict, Any, List, Optional, Tuple
+from dataclasses import dataclass, asdict
+
+
+class AgentRole(str, Enum):
+    PROFILER = "Profiler"
+    VALIDATOR = "Validator"
+    REMEDIATOR = "Remediator"
+    REVIEWER = "Reviewer"
+
+
+@dataclass
+class RoleHandoff:
+    handoff_id: str
+    from_role: str
+    to_role: str
+    timestamp: str
+    contract_goal: str
+    inputs_passed: Dict[str, Any]
+    outputs_produced: Dict[str, Any]
+    status: str = "Completed"
+
+    def to_dict(self) -> Dict[str, Any]:
+        return asdict(self)
+
+
+class RoleOrchestrator:
+    """
+    Coordinates the 4-role multi-agent workflow and records audit logs.
+    """
+
+    def __init__(self, run_id: str):
+        self.run_id = run_id
+        self.handoffs: List[RoleHandoff] = []
+
+    def record_handoff(
+        self,
+        from_role: AgentRole,
+        to_role: AgentRole,
+        contract_goal: str,
+        inputs: Dict[str, Any],
+        outputs: Dict[str, Any]
+    ) -> RoleHandoff:
+        now = datetime.now(timezone.utc).isoformat()
+        h_id = f"HDF-{len(self.handoffs) + 1:02d}"
+        handoff = RoleHandoff(
+            handoff_id=h_id,
+            from_role=from_role.value,
+            to_role=to_role.value,
+            timestamp=now,
+            contract_goal=contract_goal,
+            inputs_passed=inputs,
+            outputs_produced=outputs,
+            status="Completed"
+        )
+        self.handoffs.append(handoff)
+        return handoff
+
+    def write_artifacts(self, run_dir: Path) -> Tuple[Path, Path]:
+        run_dir.mkdir(parents=True, exist_ok=True)
+        audit_file = run_dir / "role-audit.json"
+        handoff_file = run_dir / "agent-handoffs.md"
+
+        # 1. role-audit.json
+        with open(audit_file, "w", encoding="utf-8") as f:
+            json.dump([h.to_dict() for h in self.handoffs], f, indent=2)
+
+        # 2. agent-handoffs.md
+        lines = [
+            "# TorusGuard v0.7.0 Multi-Agent Role Handoff & Governance Trail",
+            f"\n**Run Identifier:** `{self.run_id}`",
+            f"**Total Role Handoffs:** `{len(self.handoffs)}`",
+            "\n---",
+            "\n## 👥 Multi-Agent Responsibility Matrix\n",
+            "| Agent Role | Explicit Authority & Purpose | Forbidden Actions |",
+            "|---|---|---|",
+            "| **🔍 Profiler** | Detects tech stack, framework versions, route ASTs, and storage boundaries. | Cannot execute active probes or generate code patches. |",
+            "| **⚡ Validator** | Executes authorized, bounded runtime HTTP/browser probes; gathers evidence. | Cannot apply code changes or override reviewer policies. |",
+            "| **🛠️ Remediator** | Formulates minimal bounded remediation bundles enriched with runtime context. | Cannot dispatch runtime network probes or declare final sign-off. |",
+            "| **🛡️ Reviewer** | Verifies technical evidence, reviews safety decisions, and issues sign-off. | Does not execute probes or author patches. |",
+            "\n---",
+            "\n## 🔄 Chronological Handoff Log\n",
+            "| Handoff ID | From Role | To Role | Contract Goal | Timestamp | Status |",
+            "|---|---|---|---|---|:---:|",
+        ]
+        for h in self.handoffs:
+            lines.append(
+                f"| `{h.handoff_id}` | **{h.from_role}** | **{h.to_role}** | {h.contract_goal} | `{h.timestamp}` | ✅ {h.status} |"
+            )
+
+        with open(handoff_file, "w", encoding="utf-8") as f:
+            f.write("\n".join(lines) + "\n")
+
+        return audit_file, handoff_file

package/skills/torusguard/payload/core/ast_walker.py

@@ -0,0 +1,283 @@
+"""
+TorusGuard Polyglot AST Walker
+Extracts structured primitives (FunctionCall, Assignment, ImportStatement, StringLiteral)
+from Tree-sitter AST nodes and fallback parsers across polyglot languages.
+"""
+
+from dataclasses import dataclass, field, asdict
+from typing import List, Dict, Any, Optional
+
+
+@dataclass
+class FunctionCall:
+    name: str
+    full_call: str
+    arguments: List[str]
+    line_number: int
+    column: int
+    receiver: Optional[str] = None
+
+    def to_dict(self) -> Dict[str, Any]:
+        return asdict(self)
+
+
+@dataclass
+class Assignment:
+    target: str
+    value_expression: str
+    line_number: int
+    column: int
+
+    def to_dict(self) -> Dict[str, Any]:
+        return asdict(self)
+
+
+@dataclass
+class ImportStatement:
+    module: str
+    imported_names: List[str]
+    alias_map: Dict[str, str] = field(default_factory=dict)
+    line_number: int = 0
+    is_relative: bool = False
+
+    def to_dict(self) -> Dict[str, Any]:
+        return asdict(self)
+
+
+@dataclass
+class StringLiteral:
+    value: str
+    line_number: int
+    column: int
+
+    def to_dict(self) -> Dict[str, Any]:
+        return asdict(self)
+
+
+class TreeSitterWalker:
+    """
+    Traverses Tree-sitter parse trees and extracts normalized language constructs.
+    """
+
+    @staticmethod
+    def extract_function_calls(root_node, source_bytes: bytes) -> List[FunctionCall]:
+        calls: List[FunctionCall] = []
+
+        def visit(node):
+            # Python: call -> function, arguments
+            # JS/TS: call_expression -> function, arguments
+            # Go: call_expression -> function, argument_list
+            if node.type in ("call", "call_expression"):
+                func_child = node.child_by_field_name("function")
+                args_child = node.child_by_field_name("arguments") or node.child_by_field_name("argument_list")
+
+                func_name = ""
+                receiver = None
+                if func_child:
+                    func_text = source_bytes[func_child.start_byte:func_child.end_byte].decode("utf-8", errors="replace")
+                    func_name = func_text
+                    if "." in func_text:
+                        parts = func_text.split(".")
+                        receiver = ".".join(parts[:-1])
+
+                args_list: List[str] = []
+                if args_child:
+                    for ch in args_child.children:
+                        if ch.type not in (",", "(", ")"):
+                            arg_text = source_bytes[ch.start_byte:ch.end_byte].decode("utf-8", errors="replace").strip()
+                            if arg_text:
+                                args_list.append(arg_text)
+
+                line_num = node.start_point[0] + 1
+                col = node.start_point[1]
+                call_text = source_bytes[node.start_byte:node.end_byte].decode("utf-8", errors="replace")
+
+                calls.append(FunctionCall(
+                    name=func_name,
+                    full_call=call_text,
+                    arguments=args_list,
+                    line_number=line_num,
+                    column=col,
+                    receiver=receiver
+                ))
+
+            for ch in node.children:
+                visit(ch)
+
+        visit(root_node)
+        return calls
+
+    @staticmethod
+    def extract_assignments(root_node, source_bytes: bytes) -> List[Assignment]:
+        assignments: List[Assignment] = []
+
+        def visit(node):
+            # Python: assignment -> left, right
+            # JS/TS: variable_declarator (name, value), assignment_expression (left, right)
+            # Go: short_var_declaration (left, right), assignment_statement (left, right)
+            target = ""
+            val_expr = ""
+            line_num = node.start_point[0] + 1
+            col = node.start_point[1]
+
+            if node.type == "assignment":  # Python
+                left = node.child_by_field_name("left")
+                right = node.child_by_field_name("right")
+                if left and right:
+                    target = source_bytes[left.start_byte:left.end_byte].decode("utf-8", errors="replace").strip()
+                    val_expr = source_bytes[right.start_byte:right.end_byte].decode("utf-8", errors="replace").strip()
+                    assignments.append(Assignment(target=target, value_expression=val_expr, line_number=line_num, column=col))
+
+            elif node.type == "variable_declarator":  # JS/TS
+                name = node.child_by_field_name("name")
+                val = node.child_by_field_name("value")
+                if name:
+                    target = source_bytes[name.start_byte:name.end_byte].decode("utf-8", errors="replace").strip()
+                    if val:
+                        val_expr = source_bytes[val.start_byte:val.end_byte].decode("utf-8", errors="replace").strip()
+                    assignments.append(Assignment(target=target, value_expression=val_expr, line_number=line_num, column=col))
+
+            elif node.type == "assignment_expression":  # JS/TS
+                left = node.child_by_field_name("left")
+                right = node.child_by_field_name("right")
+                if left and right:
+                    target = source_bytes[left.start_byte:left.end_byte].decode("utf-8", errors="replace").strip()
+                    val_expr = source_bytes[right.start_byte:right.end_byte].decode("utf-8", errors="replace").strip()
+                    assignments.append(Assignment(target=target, value_expression=val_expr, line_number=line_num, column=col))
+
+            elif node.type in ("short_var_declaration", "assignment_statement"):  # Go
+                left = node.child_by_field_name("left")
+                right = node.child_by_field_name("right")
+                if left and right:
+                    target = source_bytes[left.start_byte:left.end_byte].decode("utf-8", errors="replace").strip()
+                    val_expr = source_bytes[right.start_byte:right.end_byte].decode("utf-8", errors="replace").strip()
+                    assignments.append(Assignment(target=target, value_expression=val_expr, line_number=line_num, column=col))
+
+            for ch in node.children:
+                visit(ch)
+
+        visit(root_node)
+        return assignments
+
+    @staticmethod
+    def extract_imports(root_node, source_bytes: bytes, language: str) -> List[ImportStatement]:
+        imports: List[ImportStatement] = []
+
+        def visit(node):
+            line_num = node.start_point[0] + 1
+
+            # Python imports
+            if node.type == "import_statement":
+                raw = source_bytes[node.start_byte:node.end_byte].decode("utf-8", errors="replace").strip()
+                # e.g. import os, sys as s
+                names = []
+                alias_map = {}
+                for ch in node.children:
+                    if ch.type == "dotted_name":
+                        mod = source_bytes[ch.start_byte:ch.end_byte].decode("utf-8", errors="replace").strip()
+                        names.append(mod)
+                    elif ch.type == "aliased_import":
+                        n = ch.child_by_field_name("name")
+                        a = ch.child_by_field_name("alias")
+                        if n and a:
+                            name_str = source_bytes[n.start_byte:n.end_byte].decode("utf-8", errors="replace").strip()
+                            alias_str = source_bytes[a.start_byte:a.end_byte].decode("utf-8", errors="replace").strip()
+                            names.append(name_str)
+                            alias_map[alias_str] = name_str
+                for n in names:
+                    imports.append(ImportStatement(module=n, imported_names=[n], alias_map=alias_map, line_number=line_num))
+
+            elif node.type == "import_from_statement":
+                # from module import a, b as c
+                module_name = ""
+                mod_child = node.child_by_field_name("module_name")
+                if mod_child:
+                    module_name = source_bytes[mod_child.start_byte:mod_child.end_byte].decode("utf-8", errors="replace").strip()
+                names = []
+                alias_map = {}
+                for ch in node.children:
+                    if ch.type == "dotted_name" and ch != mod_child:
+                        names.append(source_bytes[ch.start_byte:ch.end_byte].decode("utf-8", errors="replace").strip())
+                    elif ch.type == "aliased_import":
+                        n = ch.child_by_field_name("name")
+                        a = ch.child_by_field_name("alias")
+                        if n and a:
+                            n_str = source_bytes[n.start_byte:n.end_byte].decode("utf-8", errors="replace").strip()
+                            a_str = source_bytes[a.start_byte:a.end_byte].decode("utf-8", errors="replace").strip()
+                            names.append(n_str)
+                            alias_map[a_str] = n_str
+                is_rel = module_name.startswith(".")
+                imports.append(ImportStatement(
+                    module=module_name,
+                    imported_names=names,
+                    alias_map=alias_map,
+                    line_number=line_num,
+                    is_relative=is_rel
+                ))
+
+            # JS/TS imports: import_statement
+            elif node.type == "import_statement":
+                raw = source_bytes[node.start_byte:node.end_byte].decode("utf-8", errors="replace").strip()
+                source_child = node.child_by_field_name("source")
+                source_mod = ""
+                if source_child:
+                    source_mod = source_bytes[source_child.start_byte:source_child.end_byte].decode("utf-8", errors="replace").strip().strip("'\"")
+                names = []
+                alias_map = {}
+                for ch in node.children:
+                    if ch.type == "import_clause":
+                        clause_str = source_bytes[ch.start_byte:ch.end_byte].decode("utf-8", errors="replace").strip()
+                        names.append(clause_str)
+                is_rel = source_mod.startswith(".")
+                imports.append(ImportStatement(
+                    module=source_mod,
+                    imported_names=names,
+                    alias_map=alias_map,
+                    line_number=line_num,
+                    is_relative=is_rel
+                ))
+
+            # Go imports: import_declaration
+            elif node.type == "import_declaration":
+                for ch in node.children:
+                    if ch.type in ("import_spec", "import_spec_list"):
+                        for spec in (ch.children if ch.type == "import_spec_list" else [ch]):
+                            if spec.type == "import_spec":
+                                path_child = spec.child_by_field_name("path")
+                                if path_child:
+                                    path_str = source_bytes[path_child.start_byte:path_child.end_byte].decode("utf-8", errors="replace").strip().strip('"')
+                                    name_child = spec.child_by_field_name("name")
+                                    alias = ""
+                                    if name_child:
+                                        alias = source_bytes[name_child.start_byte:name_child.end_byte].decode("utf-8", errors="replace").strip()
+                                    alias_map = {alias: path_str} if alias else {}
+                                    imports.append(ImportStatement(
+                                        module=path_str,
+                                        imported_names=[path_str],
+                                        alias_map=alias_map,
+                                        line_number=spec.start_point[0] + 1
+                                    ))
+
+            for ch in node.children:
+                visit(ch)
+
+        visit(root_node)
+        return imports
+
+    @staticmethod
+    def extract_string_literals(root_node, source_bytes: bytes) -> List[StringLiteral]:
+        literals: List[StringLiteral] = []
+
+        def visit(node):
+            if node.type in ("string", "string_literal", "raw_string_literal", "template_string"):
+                text = source_bytes[node.start_byte:node.end_byte].decode("utf-8", errors="replace")
+                literals.append(StringLiteral(
+                    value=text.strip("\"'`"),
+                    line_number=node.start_point[0] + 1,
+                    column=node.start_point[1]
+                ))
+            for ch in node.children:
+                visit(ch)
+
+        visit(root_node)
+        return literals