foundry-mcp 0.8.22__py3-none-any.whl

foundry_mcp/tools/unified/task.py

@@ -0,0 +1,3846 @@
+"""Unified task router with validation, pagination, and shared delegates."""
+
+from __future__ import annotations
+
+import logging
+import re
+import time
+from dataclasses import asdict
+from pathlib import Path
+from typing import Any, Dict, List, Optional, Tuple
+
+from mcp.server.fastmcp import FastMCP
+
+from foundry_mcp.config import ServerConfig
+from foundry_mcp.core.context import generate_correlation_id, get_correlation_id
+from foundry_mcp.core.naming import canonical_tool
+from foundry_mcp.core.observability import get_metrics, mcp_tool
+from foundry_mcp.core.pagination import (
+    CursorError,
+    decode_cursor,
+    encode_cursor,
+    normalize_page_size,
+    paginated_response,
+)
+from foundry_mcp.core.progress import (
+    get_progress_summary,
+    list_phases,
+    sync_computed_fields,
+    update_parent_status,
+)
+from foundry_mcp.core.responses import (
+    ErrorCode,
+    ErrorType,
+    error_response,
+    success_response,
+)
+from foundry_mcp.core.spec import find_specs_directory, load_spec, save_spec
+from foundry_mcp.core.journal import (
+    add_journal_entry,
+    get_blocker_info,
+    list_blocked_tasks,
+    mark_blocked,
+    unblock as unblock_task,
+    update_task_status,
+)
+from foundry_mcp.core.task import (
+    add_task,
+    batch_update_tasks,
+    check_dependencies,
+    get_next_task,
+    manage_task_dependency,
+    move_task,
+    prepare_task as core_prepare_task,
+    remove_task,
+    REQUIREMENT_TYPES,
+    update_estimate,
+    update_task_metadata,
+    update_task_requirements,
+)
+from foundry_mcp.core.batch_operations import (
+    prepare_batch_context,
+    start_batch,
+    complete_batch,
+    reset_batch,
+    DEFAULT_MAX_TASKS,
+    DEFAULT_TOKEN_BUDGET,
+    STALE_TASK_THRESHOLD_HOURS,
+)
+from foundry_mcp.cli.context import (
+    AutonomousSession,
+    get_context_tracker,
+)
+from foundry_mcp.core.validation import (
+    VALID_VERIFICATION_TYPES,
+    VERIFICATION_TYPE_MAPPING,
+)
+from foundry_mcp.tools.unified.router import (
+    ActionDefinition,
+    ActionRouter,
+    ActionRouterError,
+)
+
+logger = logging.getLogger(__name__)
+_metrics = get_metrics()
+
+_TASK_DEFAULT_PAGE_SIZE = 25
+_TASK_MAX_PAGE_SIZE = 100
+_TASK_WARNING_THRESHOLD = 75
+_ALLOWED_STATUS = {"pending", "in_progress", "completed", "blocked"}
+
+
+def _request_id() -> str:
+    return get_correlation_id() or generate_correlation_id(prefix="task")
+
+
+def _metric(action: str) -> str:
+    return f"unified_tools.task.{action.replace('-', '_')}"
+
+
+def _specs_dir_missing_error(request_id: str) -> dict:
+    return asdict(
+        error_response(
+            "No specs directory found. Use --specs-dir or set SDD_SPECS_DIR.",
+            error_code=ErrorCode.NOT_FOUND,
+            error_type=ErrorType.NOT_FOUND,
+            remediation="Set SDD_SPECS_DIR or invoke with --specs-dir",
+            request_id=request_id,
+        )
+    )
+
+
+def _validation_error(
+    *,
+    field: str,
+    action: str,
+    message: str,
+    request_id: str,
+    code: ErrorCode = ErrorCode.MISSING_REQUIRED,
+    remediation: Optional[str] = None,
+) -> dict:
+    effective_remediation = remediation or f"Provide a valid '{field}' value"
+    return asdict(
+        error_response(
+            f"Invalid field '{field}' for task.{action}: {message}",
+            error_code=code,
+            error_type=ErrorType.VALIDATION,
+            remediation=effective_remediation,
+            details={"field": field, "action": f"task.{action}"},
+            request_id=request_id,
+        )
+    )
+
+
+def _resolve_specs_dir(
+    config: ServerConfig, workspace: Optional[str]
+) -> Optional[Path]:
+    try:
+        if workspace:
+            return find_specs_directory(workspace)
+
+        candidate = getattr(config, "specs_dir", None)
+        if isinstance(candidate, Path):
+            return candidate
+        if isinstance(candidate, str) and candidate.strip():
+            return Path(candidate)
+
+        return find_specs_directory()
+    except Exception:  # pragma: no cover - defensive guard
+        logger.exception(
+            "Failed to resolve specs directory", extra={"workspace": workspace}
+        )
+        return None
+
+
+def _load_spec_data(
+    spec_id: str, specs_dir: Optional[Path], request_id: str
+) -> Tuple[Optional[Dict[str, Any]], Optional[dict]]:
+    if specs_dir is None:
+        return None, _specs_dir_missing_error(request_id)
+
+    spec_data = load_spec(spec_id, specs_dir)
+    if spec_data is None:
+        return None, asdict(
+            error_response(
+                f"Spec not found: {spec_id}",
+                error_code=ErrorCode.SPEC_NOT_FOUND,
+                error_type=ErrorType.NOT_FOUND,
+                remediation='Verify the spec ID via spec(action="list")',
+                request_id=request_id,
+            )
+        )
+    return spec_data, None
+
+
+def _attach_meta(
+    response: dict,
+    *,
+    request_id: str,
+    duration_ms: Optional[float] = None,
+    warnings: Optional[List[str]] = None,
+) -> dict:
+    meta = response.setdefault("meta", {"version": "response-v2"})
+    meta["request_id"] = request_id
+    if warnings:
+        existing = list(meta.get("warnings") or [])
+        existing.extend(warnings)
+        meta["warnings"] = existing
+    if duration_ms is not None:
+        telemetry = dict(meta.get("telemetry") or {})
+        telemetry["duration_ms"] = round(duration_ms, 2)
+        meta["telemetry"] = telemetry
+    return response
+
+
+def _filter_hierarchy(
+    hierarchy: Dict[str, Any],
+    max_depth: int,
+    include_metadata: bool,
+) -> Dict[str, Any]:
+    result: Dict[str, Any] = {}
+
+    for node_id, node_data in hierarchy.items():
+        node_depth = node_id.count("-") if node_id != "spec-root" else 0
+        if max_depth > 0 and node_depth > max_depth:
+            continue
+
+        filtered_node: Dict[str, Any] = {
+            "type": node_data.get("type"),
+            "title": node_data.get("title"),
+            "status": node_data.get("status"),
+        }
+        if "children" in node_data:
+            filtered_node["children"] = node_data["children"]
+        if "parent" in node_data:
+            filtered_node["parent"] = node_data["parent"]
+
+        if include_metadata:
+            if "metadata" in node_data:
+                filtered_node["metadata"] = node_data["metadata"]
+            if "dependencies" in node_data:
+                filtered_node["dependencies"] = node_data["dependencies"]
+
+        result[node_id] = filtered_node
+
+    return result
+
+
+def _pagination_warnings(total_count: int, has_more: bool) -> List[str]:
+    warnings: List[str] = []
+    if total_count > _TASK_WARNING_THRESHOLD:
+        warnings.append(
+            f"{total_count} results returned; consider using pagination to limit payload size."
+        )
+    if has_more:
+        warnings.append("Additional results available. Follow the cursor to continue.")
+    return warnings
+
+
+def _handle_prepare(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "prepare"
+    spec_id = payload.get("spec_id")
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    task_id = payload.get("task_id")
+    if task_id is not None and (not isinstance(task_id, str) or not task_id.strip()):
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="task_id must be a non-empty string",
+            request_id=request_id,
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+    result = core_prepare_task(
+        spec_id=spec_id.strip(), specs_dir=specs_dir, task_id=task_id
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return _attach_meta(result, request_id=request_id, duration_ms=elapsed_ms)
+
+
+def _handle_prepare_batch(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    """
+    Handle prepare-batch action for parallel task execution.
+
+    Returns multiple independent tasks with context for parallel implementation.
+    """
+    request_id = _request_id()
+    action = "prepare-batch"
+    spec_id = payload.get("spec_id")
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+
+    # Optional parameters with defaults
+    max_tasks = payload.get("max_tasks", DEFAULT_MAX_TASKS)
+    if not isinstance(max_tasks, int) or max_tasks < 1:
+        return _validation_error(
+            field="max_tasks",
+            action=action,
+            message="max_tasks must be a positive integer",
+            request_id=request_id,
+            code=ErrorCode.VALIDATION_ERROR,
+        )
+
+    token_budget = payload.get("token_budget", DEFAULT_TOKEN_BUDGET)
+    if not isinstance(token_budget, int) or token_budget < 1000:
+        return _validation_error(
+            field="token_budget",
+            action=action,
+            message="token_budget must be an integer >= 1000",
+            request_id=request_id,
+            code=ErrorCode.VALIDATION_ERROR,
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+    result, error = prepare_batch_context(
+        spec_id=spec_id.strip(),
+        max_tasks=max_tasks,
+        token_budget=token_budget,
+        specs_dir=specs_dir,
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error:
+        _metrics.counter(_metric(action), labels={"status": "error"})
+        return asdict(
+            error_response(
+                error,
+                error_code=ErrorCode.OPERATION_FAILED,
+                error_type=ErrorType.VALIDATION,
+                request_id=request_id,
+            )
+        )
+
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+
+    # Build response with batch context
+    response = success_response(
+        spec_id=spec_id.strip(),
+        tasks=result.get("tasks", []),
+        task_count=result.get("task_count", 0),
+        spec_complete=result.get("spec_complete", False),
+        all_blocked=result.get("all_blocked", False),
+        stale_tasks=result.get("stale_tasks", []),
+        dependency_graph=result.get("dependency_graph", {}),
+        token_estimate=result.get("token_estimate", 0),
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+
+    warnings = result.get("warnings", [])
+    return _attach_meta(
+        asdict(response),
+        request_id=request_id,
+        duration_ms=elapsed_ms,
+        warnings=warnings if warnings else None,
+    )
+
+
+def _handle_start_batch(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    """
+    Handle start-batch action for atomically starting multiple tasks.
+
+    Validates all tasks can be started before making any changes.
+    """
+    request_id = _request_id()
+    action = "start-batch"
+    spec_id = payload.get("spec_id")
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+
+    task_ids = payload.get("task_ids")
+    if not isinstance(task_ids, list) or not task_ids:
+        return _validation_error(
+            field="task_ids",
+            action=action,
+            message="Provide a non-empty list of task IDs",
+            request_id=request_id,
+        )
+
+    # Validate all task_ids are strings
+    for i, tid in enumerate(task_ids):
+        if not isinstance(tid, str) or not tid.strip():
+            return _validation_error(
+                field=f"task_ids[{i}]",
+                action=action,
+                message="Each task ID must be a non-empty string",
+                request_id=request_id,
+                code=ErrorCode.VALIDATION_ERROR,
+            )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+    result, error = start_batch(
+        spec_id=spec_id.strip(),
+        task_ids=[tid.strip() for tid in task_ids],
+        specs_dir=specs_dir,
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error:
+        _metrics.counter(_metric(action), labels={"status": "error"})
+        # Include partial results in error response
+        return asdict(
+            error_response(
+                error,
+                error_code=ErrorCode.OPERATION_FAILED,
+                error_type=ErrorType.VALIDATION,
+                request_id=request_id,
+                details=result if result else None,
+            )
+        )
+
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+
+    response = success_response(
+        spec_id=spec_id.strip(),
+        started=result.get("started", []),
+        started_count=result.get("started_count", 0),
+        started_at=result.get("started_at"),
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    return _attach_meta(asdict(response), request_id=request_id, duration_ms=elapsed_ms)
+
+
+def _handle_complete_batch(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    """Handle complete-batch action for completing multiple tasks with partial failure support."""
+    request_id = _request_id()
+    action = "complete-batch"
+    spec_id = payload.get("spec_id")
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(field="spec_id", action=action, message="Provide a non-empty spec identifier", request_id=request_id)
+
+    completions = payload.get("completions")
+    if not isinstance(completions, list) or not completions:
+        return _validation_error(field="completions", action=action, message="Provide a non-empty list of completions", request_id=request_id)
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+    result, error = complete_batch(spec_id=spec_id.strip(), completions=completions, specs_dir=specs_dir)
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error:
+        _metrics.counter(_metric(action), labels={"status": "error"})
+        return asdict(error_response(error, error_code=ErrorCode.OPERATION_FAILED, error_type=ErrorType.VALIDATION, request_id=request_id, details=result if result else None))
+
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+
+    response = success_response(
+        spec_id=spec_id.strip(),
+        results=result.get("results", {}),
+        completed_count=result.get("completed_count", 0),
+        failed_count=result.get("failed_count", 0),
+        total_processed=result.get("total_processed", 0),
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    return _attach_meta(asdict(response), request_id=request_id, duration_ms=elapsed_ms)
+
+
+def _handle_reset_batch(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    """
+    Handle reset-batch action for resetting stale or specified in_progress tasks.
+
+    Resets tasks back to pending status and clears started_at timestamp.
+    If task_ids not provided, finds stale tasks automatically based on threshold.
+    """
+    request_id = _request_id()
+    action = "reset-batch"
+    spec_id = payload.get("spec_id")
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+
+    # Optional: specific task IDs to reset
+    task_ids = payload.get("task_ids")
+    if task_ids is not None:
+        if not isinstance(task_ids, list):
+            return _validation_error(
+                field="task_ids",
+                action=action,
+                message="task_ids must be a list of strings",
+                request_id=request_id,
+            )
+        # Validate all task_ids are strings
+        for i, tid in enumerate(task_ids):
+            if not isinstance(tid, str) or not tid.strip():
+                return _validation_error(
+                    field=f"task_ids[{i}]",
+                    action=action,
+                    message="Each task ID must be a non-empty string",
+                    request_id=request_id,
+                    code=ErrorCode.VALIDATION_ERROR,
+                )
+        task_ids = [tid.strip() for tid in task_ids]
+
+    # Optional: threshold in hours for stale detection
+    threshold_hours = payload.get("threshold_hours", STALE_TASK_THRESHOLD_HOURS)
+    if not isinstance(threshold_hours, (int, float)) or threshold_hours <= 0:
+        return _validation_error(
+            field="threshold_hours",
+            action=action,
+            message="threshold_hours must be a positive number",
+            request_id=request_id,
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+    result, error = reset_batch(
+        spec_id=spec_id.strip(),
+        task_ids=task_ids,
+        threshold_hours=float(threshold_hours),
+        specs_dir=specs_dir,
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error:
+        _metrics.counter(_metric(action), labels={"status": "error"})
+        return asdict(
+            error_response(
+                error,
+                error_code=ErrorCode.OPERATION_FAILED,
+                error_type=ErrorType.VALIDATION,
+                request_id=request_id,
+                details=result if result else None,
+            )
+        )
+
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+
+    response = success_response(
+        spec_id=spec_id.strip(),
+        reset=result.get("reset", []),
+        reset_count=result.get("reset_count", 0),
+        errors=result.get("errors"),
+        message=result.get("message"),
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    return _attach_meta(asdict(response), request_id=request_id, duration_ms=elapsed_ms)
+
+
+def _handle_next(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "next"
+    spec_id = payload.get("spec_id")
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None  # narrow Optional
+
+    start = time.perf_counter()
+    next_task = get_next_task(spec_data)
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    telemetry = {"duration_ms": round(elapsed_ms, 2)}
+
+    if next_task:
+        task_id, task_data = next_task
+        response = success_response(
+            spec_id=spec_id.strip(),
+            found=True,
+            task_id=task_id,
+            title=task_data.get("title", ""),
+            type=task_data.get("type", "task"),
+            status=task_data.get("status", "pending"),
+            metadata=task_data.get("metadata", {}),
+            request_id=request_id,
+            telemetry=telemetry,
+        )
+    else:
+        hierarchy = spec_data.get("hierarchy", {})
+        all_tasks = [
+            node
+            for node in hierarchy.values()
+            if node.get("type") in {"task", "subtask", "verify"}
+        ]
+        completed = sum(1 for node in all_tasks if node.get("status") == "completed")
+        pending = sum(1 for node in all_tasks if node.get("status") == "pending")
+        response = success_response(
+            spec_id=spec_id.strip(),
+            found=False,
+            spec_complete=pending == 0 and completed > 0,
+            message="All tasks completed"
+            if pending == 0 and completed > 0
+            else "No actionable tasks (tasks may be blocked)",
+            request_id=request_id,
+            telemetry=telemetry,
+        )
+
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_info(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "info"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    task = spec_data.get("hierarchy", {}).get(task_id.strip())
+    if task is None:
+        return asdict(
+            error_response(
+                f"Task not found: {task_id.strip()}",
+                error_code=ErrorCode.TASK_NOT_FOUND,
+                error_type=ErrorType.NOT_FOUND,
+                remediation="Verify the task ID exists in the hierarchy",
+                request_id=request_id,
+            )
+        )
+
+    response = success_response(
+        spec_id=spec_id.strip(),
+        task_id=task_id.strip(),
+        task=task,
+        request_id=request_id,
+    )
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_check_deps(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "check-deps"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    start = time.perf_counter()
+    deps = check_dependencies(spec_data, task_id.strip())
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    response = success_response(
+        **deps,
+        spec_id=spec_id.strip(),
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_progress(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "progress"
+    spec_id = payload.get("spec_id")
+    node_id = payload.get("node_id", "spec-root")
+    include_phases = payload.get("include_phases", True)
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(node_id, str) or not node_id.strip():
+        return _validation_error(
+            field="node_id",
+            action=action,
+            message="Provide a non-empty node identifier",
+            request_id=request_id,
+        )
+    if not isinstance(include_phases, bool):
+        return _validation_error(
+            field="include_phases",
+            action=action,
+            message="Expected a boolean value",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    progress = get_progress_summary(spec_data, node_id.strip())
+    if include_phases:
+        progress["phases"] = list_phases(spec_data)
+
+    response = success_response(
+        **progress,
+        request_id=request_id,
+    )
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_list(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "list"
+    spec_id = payload.get("spec_id")
+    status_filter = payload.get("status_filter")
+    include_completed = payload.get("include_completed", True)
+    limit = payload.get("limit")
+    cursor = payload.get("cursor")
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if status_filter is not None:
+        if not isinstance(status_filter, str) or status_filter not in _ALLOWED_STATUS:
+            return _validation_error(
+                field="status_filter",
+                action=action,
+                message=f"Status must be one of: {sorted(_ALLOWED_STATUS)}",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+    if not isinstance(include_completed, bool):
+        return _validation_error(
+            field="include_completed",
+            action=action,
+            message="Expected a boolean value",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    page_size = normalize_page_size(
+        limit,
+        default=_TASK_DEFAULT_PAGE_SIZE,
+        maximum=_TASK_MAX_PAGE_SIZE,
+    )
+
+    start_after_id = None
+    if cursor:
+        try:
+            cursor_data = decode_cursor(cursor)
+            start_after_id = cursor_data.get("last_id")
+        except CursorError as exc:
+            return asdict(
+                error_response(
+                    f"Invalid cursor: {exc.reason or exc}",
+                    error_code=ErrorCode.INVALID_FORMAT,
+                    error_type=ErrorType.VALIDATION,
+                    request_id=request_id,
+                )
+            )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    start = time.perf_counter()
+    hierarchy = spec_data.get("hierarchy", {})
+    tasks: List[Dict[str, Any]] = []
+    for node_id, node in hierarchy.items():
+        if node.get("type") not in {"task", "subtask", "verify"}:
+            continue
+        status = node.get("status", "pending")
+        if status_filter and status != status_filter:
+            continue
+        if not include_completed and status == "completed":
+            continue
+        tasks.append(
+            {
+                "id": node_id,
+                "title": node.get("title", "Untitled"),
+                "type": node.get("type", "task"),
+                "status": status,
+                "icon": node.get("icon"),
+                "file_path": node.get("metadata", {}).get("file_path"),
+                "parent": node.get("parent"),
+            }
+        )
+
+    tasks.sort(key=lambda item: item.get("id", ""))
+    total_count = len(tasks)
+
+    if start_after_id:
+        try:
+            start_index = next(
+                i for i, task in enumerate(tasks) if task.get("id") == start_after_id
+            )
+            tasks = tasks[start_index + 1 :]
+        except StopIteration:
+            pass
+
+    page_tasks = tasks[: page_size + 1]
+    has_more = len(page_tasks) > page_size
+    if has_more:
+        page_tasks = page_tasks[:page_size]
+
+    next_cursor = None
+    if has_more and page_tasks:
+        next_cursor = encode_cursor({"last_id": page_tasks[-1].get("id")})
+
+    _ = (time.perf_counter() - start) * 1000  # timing placeholder
+    warnings = _pagination_warnings(total_count, has_more)
+    response = paginated_response(
+        data={
+            "spec_id": spec_id.strip(),
+            "tasks": page_tasks,
+            "count": len(page_tasks),
+        },
+        cursor=next_cursor,
+        has_more=has_more,
+        page_size=page_size,
+        total_count=total_count,
+        warnings=warnings or None,
+        request_id=request_id,
+    )
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return response
+
+
+def _handle_query(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "query"
+    spec_id = payload.get("spec_id")
+    status = payload.get("status")
+    parent = payload.get("parent")
+    limit = payload.get("limit")
+    cursor = payload.get("cursor")
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if status is not None:
+        if not isinstance(status, str) or status not in _ALLOWED_STATUS:
+            return _validation_error(
+                field="status",
+                action=action,
+                message=f"Status must be one of: {sorted(_ALLOWED_STATUS)}",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+    if parent is not None and (not isinstance(parent, str) or not parent.strip()):
+        return _validation_error(
+            field="parent",
+            action=action,
+            message="Parent must be a non-empty string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    page_size = normalize_page_size(
+        limit,
+        default=_TASK_DEFAULT_PAGE_SIZE,
+        maximum=_TASK_MAX_PAGE_SIZE,
+    )
+
+    start_after_id = None
+    if cursor:
+        try:
+            cursor_data = decode_cursor(cursor)
+            start_after_id = cursor_data.get("last_id")
+        except CursorError as exc:
+            return asdict(
+                error_response(
+                    f"Invalid cursor: {exc.reason or exc}",
+                    error_code=ErrorCode.INVALID_FORMAT,
+                    error_type=ErrorType.VALIDATION,
+                    request_id=request_id,
+                )
+            )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    start = time.perf_counter()
+    hierarchy = spec_data.get("hierarchy", {})
+    tasks: List[Dict[str, Any]] = []
+    for task_id, task_data in hierarchy.items():
+        if status and task_data.get("status") != status:
+            continue
+        if parent and task_data.get("parent") != parent:
+            continue
+        tasks.append(
+            {
+                "task_id": task_id,
+                "title": task_data.get("title", task_id),
+                "status": task_data.get("status", "pending"),
+                "type": task_data.get("type", "task"),
+                "parent": task_data.get("parent"),
+            }
+        )
+
+    tasks.sort(key=lambda item: item.get("task_id", ""))
+    total_count = len(tasks)
+
+    if start_after_id:
+        try:
+            start_index = next(
+                i
+                for i, task in enumerate(tasks)
+                if task.get("task_id") == start_after_id
+            )
+            tasks = tasks[start_index + 1 :]
+        except StopIteration:
+            pass
+
+    page_tasks = tasks[: page_size + 1]
+    has_more = len(page_tasks) > page_size
+    if has_more:
+        page_tasks = page_tasks[:page_size]
+
+    next_cursor = None
+    if has_more and page_tasks:
+        next_cursor = encode_cursor({"last_id": page_tasks[-1].get("task_id")})
+
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    warnings = _pagination_warnings(total_count, has_more)
+    response = paginated_response(
+        data={
+            "spec_id": spec_id.strip(),
+            "tasks": page_tasks,
+            "count": len(page_tasks),
+        },
+        cursor=next_cursor,
+        has_more=has_more,
+        page_size=page_size,
+        total_count=total_count,
+        warnings=warnings or None,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return response
+
+
+def _handle_hierarchy(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "hierarchy"
+    spec_id = payload.get("spec_id")
+    max_depth = payload.get("max_depth", 2)
+    include_metadata = payload.get("include_metadata", False)
+    limit = payload.get("limit")
+    cursor = payload.get("cursor")
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(max_depth, int) or max_depth < 0 or max_depth > 10:
+        return _validation_error(
+            field="max_depth",
+            action=action,
+            message="max_depth must be between 0 and 10",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    if not isinstance(include_metadata, bool):
+        return _validation_error(
+            field="include_metadata",
+            action=action,
+            message="Expected a boolean value",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    page_size = normalize_page_size(
+        limit,
+        default=_TASK_DEFAULT_PAGE_SIZE,
+        maximum=_TASK_MAX_PAGE_SIZE,
+    )
+
+    start_after_id = None
+    if cursor:
+        try:
+            cursor_data = decode_cursor(cursor)
+            start_after_id = cursor_data.get("last_id")
+        except CursorError as exc:
+            return asdict(
+                error_response(
+                    f"Invalid cursor: {exc.reason or exc}",
+                    error_code=ErrorCode.INVALID_FORMAT,
+                    error_type=ErrorType.VALIDATION,
+                    request_id=request_id,
+                )
+            )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    start = time.perf_counter()
+    full_hierarchy = spec_data.get("hierarchy", {})
+    filtered = _filter_hierarchy(full_hierarchy, max_depth, include_metadata)
+    sorted_ids = sorted(filtered.keys())
+
+    if start_after_id:
+        try:
+            start_index = sorted_ids.index(start_after_id) + 1
+        except ValueError:
+            start_index = 0
+    else:
+        start_index = 0
+
+    page_ids = sorted_ids[start_index : start_index + page_size + 1]
+    has_more = len(page_ids) > page_size
+    if has_more:
+        page_ids = page_ids[:page_size]
+
+    hierarchy_page = {node_id: filtered[node_id] for node_id in page_ids}
+    next_cursor = None
+    if has_more and page_ids:
+        next_cursor = encode_cursor({"last_id": page_ids[-1]})
+
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    warnings = _pagination_warnings(len(filtered), has_more)
+    response = paginated_response(
+        data={
+            "spec_id": spec_id.strip(),
+            "hierarchy": hierarchy_page,
+            "node_count": len(hierarchy_page),
+            "total_nodes": len(filtered),
+            "filters_applied": {
+                "max_depth": max_depth,
+                "include_metadata": include_metadata,
+            },
+        },
+        cursor=next_cursor,
+        has_more=has_more,
+        page_size=page_size,
+        total_count=len(filtered),
+        warnings=warnings or None,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return response
+
+
+def _handle_update_status(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "update-status"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+    status = payload.get("status")
+    note = payload.get("note")
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+    if not isinstance(status, str) or status not in _ALLOWED_STATUS:
+        return _validation_error(
+            field="status",
+            action=action,
+            message=f"Status must be one of: {sorted(_ALLOWED_STATUS)}",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    if note is not None and (not isinstance(note, str) or not note.strip()):
+        return _validation_error(
+            field="note",
+            action=action,
+            message="note must be a non-empty string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    hierarchy = spec_data.get("hierarchy", {})
+    task_key = task_id.strip()
+    if task_key not in hierarchy:
+        return asdict(
+            error_response(
+                f"Task not found: {task_key}",
+                error_code=ErrorCode.TASK_NOT_FOUND,
+                error_type=ErrorType.NOT_FOUND,
+                remediation="Verify the task ID exists in the hierarchy",
+                request_id=request_id,
+            )
+        )
+
+    start = time.perf_counter()
+    updated = update_task_status(spec_data, task_key, status, note=None)
+    if not updated:
+        return asdict(
+            error_response(
+                f"Failed to update task status for {task_key}",
+                error_code=ErrorCode.CONFLICT,
+                error_type=ErrorType.CONFLICT,
+                remediation="Confirm the task exists and the status is valid",
+                request_id=request_id,
+            )
+        )
+
+    update_parent_status(spec_data, task_key)
+
+    if note:
+        add_journal_entry(
+            spec_data,
+            title=f"Status changed to {status}",
+            content=note,
+            entry_type="status_change",
+            task_id=task_key,
+            author="foundry-mcp",
+        )
+
+    if specs_dir is None or not save_spec(spec_id.strip(), spec_data, specs_dir):
+        return asdict(
+            error_response(
+                "Failed to save spec",
+                error_code=ErrorCode.INTERNAL_ERROR,
+                error_type=ErrorType.INTERNAL,
+                remediation="Check filesystem permissions and retry",
+                request_id=request_id,
+            )
+        )
+
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    response = success_response(
+        spec_id=spec_id.strip(),
+        task_id=task_key,
+        new_status=status,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_start(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "start"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+    note = payload.get("note")
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+    if note is not None and (not isinstance(note, str) or not note.strip()):
+        return _validation_error(
+            field="note",
+            action=action,
+            message="note must be a non-empty string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    start = time.perf_counter()
+    deps = check_dependencies(spec_data, task_id.strip())
+    if not deps.get("can_start", False):
+        blockers = [
+            b.get("title", b.get("id", ""))
+            for b in (deps.get("blocked_by") or [])
+            if isinstance(b, dict)
+        ]
+        return asdict(
+            error_response(
+                "Task is blocked by: " + ", ".join([b for b in blockers if b]),
+                error_code=ErrorCode.CONFLICT,
+                error_type=ErrorType.CONFLICT,
+                remediation="Resolve blocking tasks then retry",
+                details={"blocked_by": deps.get("blocked_by")},
+                request_id=request_id,
+            )
+        )
+
+    updated = update_task_status(spec_data, task_id.strip(), "in_progress", note=None)
+    if not updated:
+        return asdict(
+            error_response(
+                f"Failed to start task: {task_id.strip()}",
+                error_code=ErrorCode.CONFLICT,
+                error_type=ErrorType.CONFLICT,
+                remediation="Confirm the task exists and is not blocked",
+                request_id=request_id,
+            )
+        )
+
+    update_parent_status(spec_data, task_id.strip())
+    sync_computed_fields(spec_data)
+
+    if note:
+        add_journal_entry(
+            spec_data,
+            title=f"Task Started: {task_id.strip()}",
+            content=note,
+            entry_type="status_change",
+            task_id=task_id.strip(),
+            author="foundry-mcp",
+        )
+
+    if specs_dir is None or not save_spec(spec_id.strip(), spec_data, specs_dir):
+        return asdict(
+            error_response(
+                "Failed to save spec",
+                error_code=ErrorCode.INTERNAL_ERROR,
+                error_type=ErrorType.INTERNAL,
+                remediation="Check filesystem permissions and retry",
+                request_id=request_id,
+            )
+        )
+
+    task_data = spec_data.get("hierarchy", {}).get(task_id.strip(), {})
+    started_at = task_data.get("metadata", {}).get("started_at")
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    response = success_response(
+        spec_id=spec_id.strip(),
+        task_id=task_id.strip(),
+        started_at=started_at,
+        title=task_data.get("title", ""),
+        type=task_data.get("type", "task"),
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_complete(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "complete"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+    completion_note = payload.get("completion_note")
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+    if not isinstance(completion_note, str) or not completion_note.strip():
+        return _validation_error(
+            field="completion_note",
+            action=action,
+            message="Provide a non-empty completion note",
+            request_id=request_id,
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    start = time.perf_counter()
+    updated = update_task_status(spec_data, task_id.strip(), "completed", note=None)
+    if not updated:
+        return asdict(
+            error_response(
+                f"Failed to complete task: {task_id.strip()}",
+                error_code=ErrorCode.CONFLICT,
+                error_type=ErrorType.CONFLICT,
+                remediation="Confirm the task exists and is not already completed",
+                request_id=request_id,
+            )
+        )
+
+    update_parent_status(spec_data, task_id.strip())
+    sync_computed_fields(spec_data)
+
+    task_data = spec_data.get("hierarchy", {}).get(task_id.strip(), {})
+
+    # Determine if commit is suggested based on git cadence config
+    suggest_commit = False
+    commit_scope: Optional[str] = None
+    commit_message_hint: Optional[str] = None
+
+    if config.git.enabled:
+        cadence = config.git.commit_cadence
+        hierarchy = spec_data.get("hierarchy", {})
+
+        if cadence == "task":
+            suggest_commit = True
+            commit_scope = "task"
+            commit_message_hint = f"task: {task_data.get('title', task_id.strip())}"
+        elif cadence == "phase":
+            # Check if parent phase just completed
+            parent_id = task_data.get("parent")
+            if parent_id:
+                parent_data = hierarchy.get(parent_id, {})
+                # Only suggest commit if parent is a phase and is now completed
+                if (
+                    parent_data.get("type") == "phase"
+                    and parent_data.get("status") == "completed"
+                ):
+                    suggest_commit = True
+                    commit_scope = "phase"
+                    commit_message_hint = (
+                        f"phase: {parent_data.get('title', parent_id)}"
+                    )
+    add_journal_entry(
+        spec_data,
+        title=f"Task Completed: {task_data.get('title', task_id.strip())}",
+        content=completion_note,
+        entry_type="status_change",
+        task_id=task_id.strip(),
+        author="foundry-mcp",
+    )
+
+    if specs_dir is None or not save_spec(spec_id.strip(), spec_data, specs_dir):
+        return asdict(
+            error_response(
+                "Failed to save spec",
+                error_code=ErrorCode.INTERNAL_ERROR,
+                error_type=ErrorType.INTERNAL,
+                remediation="Check filesystem permissions and retry",
+                request_id=request_id,
+            )
+        )
+
+    completed_at = task_data.get("metadata", {}).get("completed_at")
+    progress = get_progress_summary(spec_data)
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    response = success_response(
+        spec_id=spec_id.strip(),
+        task_id=task_id.strip(),
+        completed_at=completed_at,
+        progress={
+            "completed_tasks": progress.get("completed_tasks", 0),
+            "total_tasks": progress.get("total_tasks", 0),
+            "percentage": progress.get("percentage", 0),
+        },
+        suggest_commit=suggest_commit,
+        commit_scope=commit_scope,
+        commit_message_hint=commit_message_hint,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_block(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "block"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+    reason = payload.get("reason")
+    blocker_type = payload.get("blocker_type", "dependency")
+    ticket = payload.get("ticket")
+
+    valid_types = {"dependency", "technical", "resource", "decision"}
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+    if not isinstance(reason, str) or not reason.strip():
+        return _validation_error(
+            field="reason",
+            action=action,
+            message="Provide a non-empty blocker reason",
+            request_id=request_id,
+        )
+    if not isinstance(blocker_type, str) or blocker_type not in valid_types:
+        return _validation_error(
+            field="blocker_type",
+            action=action,
+            message=f"blocker_type must be one of: {sorted(valid_types)}",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    if ticket is not None and not isinstance(ticket, str):
+        return _validation_error(
+            field="ticket",
+            action=action,
+            message="ticket must be a string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    start = time.perf_counter()
+    blocked = mark_blocked(
+        spec_data,
+        task_id.strip(),
+        reason.strip(),
+        blocker_type=blocker_type,
+        ticket=ticket,
+    )
+    if not blocked:
+        return asdict(
+            error_response(
+                f"Task not found: {task_id.strip()}",
+                error_code=ErrorCode.TASK_NOT_FOUND,
+                error_type=ErrorType.NOT_FOUND,
+                remediation="Verify the task ID exists in the hierarchy",
+                request_id=request_id,
+            )
+        )
+
+    add_journal_entry(
+        spec_data,
+        title=f"Task Blocked: {task_id.strip()}",
+        content=f"Blocker ({blocker_type}): {reason.strip()}"
+        + (f" [Ticket: {ticket}]" if ticket else ""),
+        entry_type="blocker",
+        task_id=task_id.strip(),
+        author="foundry-mcp",
+    )
+    sync_computed_fields(spec_data)
+
+    if specs_dir is None or not save_spec(spec_id.strip(), spec_data, specs_dir):
+        return asdict(
+            error_response(
+                "Failed to save spec",
+                error_code=ErrorCode.INTERNAL_ERROR,
+                error_type=ErrorType.INTERNAL,
+                remediation="Check filesystem permissions and retry",
+                request_id=request_id,
+            )
+        )
+
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    response = success_response(
+        spec_id=spec_id.strip(),
+        task_id=task_id.strip(),
+        blocker_type=blocker_type,
+        reason=reason.strip(),
+        ticket=ticket,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_unblock(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "unblock"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+    resolution = payload.get("resolution")
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+    if resolution is not None and (
+        not isinstance(resolution, str) or not resolution.strip()
+    ):
+        return _validation_error(
+            field="resolution",
+            action=action,
+            message="resolution must be a non-empty string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    start = time.perf_counter()
+    blocker = get_blocker_info(spec_data, task_id.strip())
+    if blocker is None:
+        return asdict(
+            error_response(
+                f"Task {task_id.strip()} is not blocked",
+                error_code=ErrorCode.CONFLICT,
+                error_type=ErrorType.CONFLICT,
+                remediation="Confirm the task is currently blocked before unblocking",
+                request_id=request_id,
+            )
+        )
+
+    unblocked = unblock_task(spec_data, task_id.strip(), resolution)
+    if not unblocked:
+        return asdict(
+            error_response(
+                f"Failed to unblock task: {task_id.strip()}",
+                error_code=ErrorCode.CONFLICT,
+                error_type=ErrorType.CONFLICT,
+                remediation="Confirm the task exists and is currently blocked",
+                request_id=request_id,
+            )
+        )
+
+    add_journal_entry(
+        spec_data,
+        title=f"Task Unblocked: {task_id.strip()}",
+        content=f"Resolved: {resolution.strip() if isinstance(resolution, str) else 'Blocker resolved'}",
+        entry_type="note",
+        task_id=task_id.strip(),
+        author="foundry-mcp",
+    )
+    sync_computed_fields(spec_data)
+
+    if specs_dir is None or not save_spec(spec_id.strip(), spec_data, specs_dir):
+        return asdict(
+            error_response(
+                "Failed to save spec",
+                error_code=ErrorCode.INTERNAL_ERROR,
+                error_type=ErrorType.INTERNAL,
+                remediation="Check filesystem permissions and retry",
+                request_id=request_id,
+            )
+        )
+
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    response = success_response(
+        spec_id=spec_id.strip(),
+        task_id=task_id.strip(),
+        previous_blocker={
+            "type": blocker.blocker_type,
+            "description": blocker.description,
+        },
+        resolution=(resolution.strip() if isinstance(resolution, str) else None)
+        or "Blocker resolved",
+        new_status="pending",
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_list_blocked(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "list-blocked"
+    spec_id = payload.get("spec_id")
+    cursor = payload.get("cursor")
+    limit = payload.get("limit")
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+
+    page_size = normalize_page_size(
+        limit,
+        default=_TASK_DEFAULT_PAGE_SIZE,
+        maximum=_TASK_MAX_PAGE_SIZE,
+    )
+
+    start_after_id = None
+    if cursor:
+        try:
+            cursor_data = decode_cursor(cursor)
+            start_after_id = cursor_data.get("last_id")
+        except CursorError as exc:
+            return asdict(
+                error_response(
+                    f"Invalid cursor: {exc.reason or exc}",
+                    error_code=ErrorCode.INVALID_FORMAT,
+                    error_type=ErrorType.VALIDATION,
+                    request_id=request_id,
+                )
+            )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    start = time.perf_counter()
+    blocked_tasks = list_blocked_tasks(spec_data)
+    blocked_tasks.sort(key=lambda entry: entry.get("task_id", ""))
+    total_count = len(blocked_tasks)
+
+    if start_after_id:
+        try:
+            start_index = next(
+                i
+                for i, entry in enumerate(blocked_tasks)
+                if entry.get("task_id") == start_after_id
+            )
+            blocked_tasks = blocked_tasks[start_index + 1 :]
+        except StopIteration:
+            pass
+
+    page_tasks = blocked_tasks[: page_size + 1]
+    has_more = len(page_tasks) > page_size
+    if has_more:
+        page_tasks = page_tasks[:page_size]
+
+    next_cursor = None
+    if has_more and page_tasks:
+        next_cursor = encode_cursor({"last_id": page_tasks[-1].get("task_id")})
+
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    warnings = _pagination_warnings(total_count, has_more)
+    response = paginated_response(
+        data={
+            "spec_id": spec_id.strip(),
+            "count": len(page_tasks),
+            "blocked_tasks": page_tasks,
+        },
+        cursor=next_cursor,
+        has_more=has_more,
+        page_size=page_size,
+        total_count=total_count,
+        warnings=warnings or None,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return response
+
+
+def _handle_add(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "add"
+    spec_id = payload.get("spec_id")
+    parent = payload.get("parent")
+    phase_id = payload.get("phase_id")  # Alias for parent
+
+    # Use phase_id as parent if parent not provided
+    if parent is None and phase_id is not None:
+        parent = phase_id
+
+    title = payload.get("title")
+    description = payload.get("description")
+    task_type = payload.get("task_type", "task")
+    estimated_hours = payload.get("estimated_hours")
+    position = payload.get("position")
+    file_path = payload.get("file_path")
+
+    # Research-specific parameters
+    research_type = payload.get("research_type")
+    blocking_mode = payload.get("blocking_mode")
+    query = payload.get("query")
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(parent, str) or not parent.strip():
+        return _validation_error(
+            field="parent",
+            action=action,
+            message="Provide a non-empty parent node identifier",
+            request_id=request_id,
+        )
+    if not isinstance(title, str) or not title.strip():
+        return _validation_error(
+            field="title",
+            action=action,
+            message="Provide a non-empty task title",
+            request_id=request_id,
+        )
+    if description is not None and not isinstance(description, str):
+        return _validation_error(
+            field="description",
+            action=action,
+            message="description must be a string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    if not isinstance(task_type, str):
+        return _validation_error(
+            field="task_type",
+            action=action,
+            message="task_type must be a string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    if estimated_hours is not None and not isinstance(estimated_hours, (int, float)):
+        return _validation_error(
+            field="estimated_hours",
+            action=action,
+            message="estimated_hours must be a number",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    if position is not None and (not isinstance(position, int) or position < 0):
+        return _validation_error(
+            field="position",
+            action=action,
+            message="position must be a non-negative integer",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    if file_path is not None and not isinstance(file_path, str):
+        return _validation_error(
+            field="file_path",
+            action=action,
+            message="file_path must be a string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    # Validate research-specific parameters when task_type is "research"
+    if task_type == "research":
+        from foundry_mcp.core.validation import VALID_RESEARCH_TYPES, RESEARCH_BLOCKING_MODES
+
+        if research_type is not None and not isinstance(research_type, str):
+            return _validation_error(
+                field="research_type",
+                action=action,
+                message="research_type must be a string",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+        if research_type and research_type not in VALID_RESEARCH_TYPES:
+            return _validation_error(
+                field="research_type",
+                action=action,
+                message=f"Must be one of: {', '.join(sorted(VALID_RESEARCH_TYPES))}",
+                request_id=request_id,
+            )
+        if blocking_mode is not None and not isinstance(blocking_mode, str):
+            return _validation_error(
+                field="blocking_mode",
+                action=action,
+                message="blocking_mode must be a string",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+        if blocking_mode and blocking_mode not in RESEARCH_BLOCKING_MODES:
+            return _validation_error(
+                field="blocking_mode",
+                action=action,
+                message=f"Must be one of: {', '.join(sorted(RESEARCH_BLOCKING_MODES))}",
+                request_id=request_id,
+            )
+        if query is not None and not isinstance(query, str):
+            return _validation_error(
+                field="query",
+                action=action,
+                message="query must be a string",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+
+    dry_run = payload.get("dry_run", False)
+    if dry_run is not None and not isinstance(dry_run, bool):
+        return _validation_error(
+            field="dry_run",
+            action=action,
+            message="dry_run must be a boolean",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    dry_run_bool = bool(dry_run)
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+    if dry_run_bool:
+        spec_data, spec_error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+        if spec_error:
+            return spec_error
+
+        hierarchy = (spec_data or {}).get("hierarchy", {})
+        parent_node = (
+            hierarchy.get(parent.strip()) if isinstance(hierarchy, dict) else None
+        )
+        if not isinstance(parent_node, dict):
+            elapsed_ms = (time.perf_counter() - start) * 1000
+            return asdict(
+                error_response(
+                    f"Parent node '{parent.strip()}' not found",
+                    error_code=ErrorCode.NOT_FOUND,
+                    error_type=ErrorType.NOT_FOUND,
+                    remediation="Verify the parent node ID exists in the specification",
+                    request_id=request_id,
+                    telemetry={"duration_ms": round(elapsed_ms, 2)},
+                )
+            )
+
+        elapsed_ms = (time.perf_counter() - start) * 1000
+        dry_run_data: Dict[str, Any] = {
+            "spec_id": spec_id.strip(),
+            "parent": parent.strip(),
+            "title": title.strip(),
+            "task_type": task_type,
+            "position": position,
+            "file_path": file_path.strip() if file_path else None,
+            "dry_run": True,
+        }
+        # Include research parameters in dry_run response
+        if task_type == "research":
+            dry_run_data["research_type"] = research_type
+            dry_run_data["blocking_mode"] = blocking_mode
+            dry_run_data["query"] = query
+        response = success_response(
+            data=dry_run_data,
+            request_id=request_id,
+            telemetry={"duration_ms": round(elapsed_ms, 2)},
+        )
+        _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+        _metrics.counter(
+            _metric(action), labels={"status": "success", "dry_run": "true"}
+        )
+        return asdict(response)
+
+    result, error = add_task(
+        spec_id=spec_id.strip(),
+        parent_id=parent.strip(),
+        title=title.strip(),
+        description=description,
+        task_type=task_type,
+        estimated_hours=float(estimated_hours) if estimated_hours is not None else None,
+        position=position,
+        file_path=file_path,
+        specs_dir=specs_dir,
+        # Research-specific parameters
+        research_type=research_type,
+        blocking_mode=blocking_mode,
+        query=query,
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error or result is None:
+        code = (
+            ErrorCode.NOT_FOUND
+            if "not found" in (error or "").lower()
+            else ErrorCode.VALIDATION_ERROR
+        )
+        err_type = (
+            ErrorType.NOT_FOUND if code == ErrorCode.NOT_FOUND else ErrorType.VALIDATION
+        )
+        return asdict(
+            error_response(
+                error or "Failed to add task",
+                error_code=code,
+                error_type=err_type,
+                remediation="Verify parent/task inputs and retry",
+                request_id=request_id,
+            )
+        )
+
+    response = success_response(
+        **result,
+        spec_id=spec_id.strip(),
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_remove(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "remove"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+    cascade = payload.get("cascade", False)
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+    if not isinstance(cascade, bool):
+        return _validation_error(
+            field="cascade",
+            action=action,
+            message="cascade must be a boolean",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    dry_run = payload.get("dry_run", False)
+    if dry_run is not None and not isinstance(dry_run, bool):
+        return _validation_error(
+            field="dry_run",
+            action=action,
+            message="dry_run must be a boolean",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    dry_run_bool = bool(dry_run)
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+    if dry_run_bool:
+        spec_data, spec_error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+        if spec_error:
+            return spec_error
+
+        hierarchy = (spec_data or {}).get("hierarchy", {})
+        node = hierarchy.get(task_id.strip()) if isinstance(hierarchy, dict) else None
+        if not isinstance(node, dict):
+            elapsed_ms = (time.perf_counter() - start) * 1000
+            return asdict(
+                error_response(
+                    f"Task '{task_id.strip()}' not found",
+                    error_code=ErrorCode.TASK_NOT_FOUND,
+                    error_type=ErrorType.NOT_FOUND,
+                    remediation="Verify the task ID exists in the specification",
+                    request_id=request_id,
+                    telemetry={"duration_ms": round(elapsed_ms, 2)},
+                )
+            )
+
+        elapsed_ms = (time.perf_counter() - start) * 1000
+        response = success_response(
+            data={
+                "spec_id": spec_id.strip(),
+                "task_id": task_id.strip(),
+                "cascade": cascade,
+                "dry_run": True,
+            },
+            request_id=request_id,
+            telemetry={"duration_ms": round(elapsed_ms, 2)},
+        )
+        _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+        _metrics.counter(
+            _metric(action), labels={"status": "success", "dry_run": "true"}
+        )
+        return asdict(response)
+
+    result, error = remove_task(
+        spec_id=spec_id.strip(),
+        task_id=task_id.strip(),
+        cascade=cascade,
+        specs_dir=specs_dir,
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error or result is None:
+        code = (
+            ErrorCode.NOT_FOUND
+            if "not found" in (error or "").lower()
+            else ErrorCode.VALIDATION_ERROR
+        )
+        err_type = (
+            ErrorType.NOT_FOUND if code == ErrorCode.NOT_FOUND else ErrorType.VALIDATION
+        )
+        return asdict(
+            error_response(
+                error or "Failed to remove task",
+                error_code=code,
+                error_type=err_type,
+                remediation="Verify the task ID and cascade flag",
+                request_id=request_id,
+            )
+        )
+
+    response = success_response(
+        **result,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_update_estimate(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "update-estimate"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+    estimated_hours = payload.get("estimated_hours")
+    complexity = payload.get("complexity")
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+    if estimated_hours is not None and not isinstance(estimated_hours, (int, float)):
+        return _validation_error(
+            field="estimated_hours",
+            action=action,
+            message="estimated_hours must be a number",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    if complexity is not None and not isinstance(complexity, str):
+        return _validation_error(
+            field="complexity",
+            action=action,
+            message="complexity must be a string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    dry_run = payload.get("dry_run", False)
+    if dry_run is not None and not isinstance(dry_run, bool):
+        return _validation_error(
+            field="dry_run",
+            action=action,
+            message="dry_run must be a boolean",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    dry_run_bool = bool(dry_run)
+
+    normalized_complexity: Optional[str] = None
+    if isinstance(complexity, str):
+        normalized_complexity = complexity.strip().lower() or None
+
+    if estimated_hours is None and normalized_complexity is None:
+        return _validation_error(
+            field="estimated_hours",
+            action=action,
+            message="Provide estimated_hours and/or complexity",
+            request_id=request_id,
+            code=ErrorCode.MISSING_REQUIRED,
+            remediation="Provide hours and/or complexity to update",
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+    if dry_run_bool:
+        spec_data, spec_error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+        if spec_error:
+            return spec_error
+
+        hierarchy = (spec_data or {}).get("hierarchy", {})
+        task = hierarchy.get(task_id.strip()) if isinstance(hierarchy, dict) else None
+        if not isinstance(task, dict):
+            return asdict(
+                error_response(
+                    f"Task '{task_id.strip()}' not found",
+                    error_code=ErrorCode.TASK_NOT_FOUND,
+                    error_type=ErrorType.NOT_FOUND,
+                    remediation="Verify the task ID exists in the specification",
+                    request_id=request_id,
+                )
+            )
+
+        metadata_candidate = task.get("metadata")
+        if isinstance(metadata_candidate, dict):
+            metadata: Dict[str, Any] = metadata_candidate
+        else:
+            metadata = {}
+        data: Dict[str, Any] = {
+            "spec_id": spec_id.strip(),
+            "task_id": task_id.strip(),
+            "dry_run": True,
+            "previous_hours": metadata.get("estimated_hours"),
+            "previous_complexity": metadata.get("complexity"),
+        }
+        if estimated_hours is not None:
+            data["hours"] = float(estimated_hours)
+        if normalized_complexity is not None:
+            data["complexity"] = normalized_complexity
+
+        elapsed_ms = (time.perf_counter() - start) * 1000
+        response = success_response(
+            data=data,
+            request_id=request_id,
+            telemetry={"duration_ms": round(elapsed_ms, 2)},
+        )
+        _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+        _metrics.counter(
+            _metric(action), labels={"status": "success", "dry_run": "true"}
+        )
+        return asdict(response)
+
+    result, error = update_estimate(
+        spec_id=spec_id.strip(),
+        task_id=task_id.strip(),
+        estimated_hours=float(estimated_hours) if estimated_hours is not None else None,
+        complexity=normalized_complexity,
+        specs_dir=specs_dir,
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error or result is None:
+        code = (
+            ErrorCode.NOT_FOUND
+            if "not found" in (error or "").lower()
+            else ErrorCode.VALIDATION_ERROR
+        )
+        err_type = (
+            ErrorType.NOT_FOUND if code == ErrorCode.NOT_FOUND else ErrorType.VALIDATION
+        )
+        return asdict(
+            error_response(
+                error or "Failed to update estimate",
+                error_code=code,
+                error_type=err_type,
+                remediation="Provide estimated_hours and/or a valid complexity",
+                request_id=request_id,
+            )
+        )
+
+    response = success_response(
+        **result,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_update_metadata(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    request_id = _request_id()
+    action = "update-metadata"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+
+    dry_run = payload.get("dry_run", False)
+    if dry_run is not None and not isinstance(dry_run, bool):
+        return _validation_error(
+            field="dry_run",
+            action=action,
+            message="dry_run must be a boolean",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    dry_run_bool = bool(dry_run)
+
+    custom_metadata = payload.get("custom_metadata")
+    if custom_metadata is not None and not isinstance(custom_metadata, dict):
+        return _validation_error(
+            field="custom_metadata",
+            action=action,
+            message="custom_metadata must be an object",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+            remediation="Provide custom_metadata as a JSON object",
+        )
+
+    acceptance_criteria = payload.get("acceptance_criteria")
+    if acceptance_criteria is not None and not isinstance(acceptance_criteria, list):
+        return _validation_error(
+            field="acceptance_criteria",
+            action=action,
+            message="acceptance_criteria must be a list of strings",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    update_fields = [
+        payload.get("title"),
+        payload.get("file_path"),
+        payload.get("description"),
+        acceptance_criteria,
+        payload.get("task_category"),
+        payload.get("actual_hours"),
+        payload.get("status_note"),
+        payload.get("verification_type"),
+        payload.get("command"),
+    ]
+    has_update = any(field is not None for field in update_fields) or bool(
+        custom_metadata
+    )
+    if not has_update:
+        return _validation_error(
+            field="title",
+            action=action,
+            message="Provide at least one field to update",
+            request_id=request_id,
+            code=ErrorCode.MISSING_REQUIRED,
+            remediation="Provide title, file_path, description, acceptance_criteria, task_category, actual_hours, status_note, verification_type, command, and/or custom_metadata",
+        )
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+    if dry_run_bool:
+        spec_data, spec_error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+        if spec_error:
+            return spec_error
+
+        hierarchy = (spec_data or {}).get("hierarchy", {})
+        task = hierarchy.get(task_id.strip()) if isinstance(hierarchy, dict) else None
+        if not isinstance(task, dict):
+            return asdict(
+                error_response(
+                    f"Task '{task_id.strip()}' not found",
+                    error_code=ErrorCode.TASK_NOT_FOUND,
+                    error_type=ErrorType.NOT_FOUND,
+                    remediation="Verify the task ID exists in the specification",
+                    request_id=request_id,
+                )
+            )
+
+        fields_updated: List[str] = []
+        if payload.get("title") is not None:
+            fields_updated.append("title")
+        if payload.get("file_path") is not None:
+            fields_updated.append("file_path")
+        if payload.get("description") is not None:
+            fields_updated.append("description")
+        if acceptance_criteria is not None:
+            fields_updated.append("acceptance_criteria")
+        if payload.get("task_category") is not None:
+            fields_updated.append("task_category")
+        if payload.get("actual_hours") is not None:
+            fields_updated.append("actual_hours")
+        if payload.get("status_note") is not None:
+            fields_updated.append("status_note")
+        if payload.get("verification_type") is not None:
+            fields_updated.append("verification_type")
+        if payload.get("command") is not None:
+            fields_updated.append("command")
+        if custom_metadata:
+            fields_updated.extend(sorted(custom_metadata.keys()))
+
+        elapsed_ms = (time.perf_counter() - start) * 1000
+        response = success_response(
+            data={
+                "spec_id": spec_id.strip(),
+                "task_id": task_id.strip(),
+                "fields_updated": fields_updated,
+                "dry_run": True,
+            },
+            request_id=request_id,
+            telemetry={"duration_ms": round(elapsed_ms, 2)},
+        )
+        _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+        _metrics.counter(
+            _metric(action), labels={"status": "success", "dry_run": "true"}
+        )
+        return asdict(response)
+
+    result, error = update_task_metadata(
+        spec_id=spec_id.strip(),
+        task_id=task_id.strip(),
+        title=payload.get("title"),
+        file_path=payload.get("file_path"),
+        description=payload.get("description"),
+        acceptance_criteria=acceptance_criteria,
+        task_category=payload.get("task_category"),
+        actual_hours=payload.get("actual_hours"),
+        status_note=payload.get("status_note"),
+        verification_type=payload.get("verification_type"),
+        command=payload.get("command"),
+        custom_metadata=custom_metadata,
+        dry_run=dry_run_bool,
+        specs_dir=specs_dir,
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error or result is None:
+        code = (
+            ErrorCode.NOT_FOUND
+            if "not found" in (error or "").lower()
+            else ErrorCode.VALIDATION_ERROR
+        )
+        err_type = (
+            ErrorType.NOT_FOUND if code == ErrorCode.NOT_FOUND else ErrorType.VALIDATION
+        )
+        return asdict(
+            error_response(
+                error or "Failed to update metadata",
+                error_code=code,
+                error_type=err_type,
+                remediation="Provide at least one metadata field to update",
+                request_id=request_id,
+            )
+        )
+
+    response = success_response(
+        **result,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_move(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    """Move a task to a new position or parent.
+
+    Supports two modes:
+    1. Reorder within parent: only specify position (new_parent=None)
+    2. Reparent to different phase/task: specify new_parent, optionally position
+
+    Updates task counts on affected parents. Prevents circular references.
+    Emits warnings for cross-phase moves that might affect dependencies.
+    """
+    request_id = _request_id()
+    action = "move"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+    new_parent = payload.get("parent")  # Target parent (phase or task ID)
+    position = payload.get("position")  # 1-based position in children list
+
+    # Validate required fields
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+
+    # Validate optional new_parent
+    if new_parent is not None and (
+        not isinstance(new_parent, str) or not new_parent.strip()
+    ):
+        return _validation_error(
+            field="parent",
+            action=action,
+            message="parent must be a non-empty string if provided",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    # Validate optional position (must be positive integer)
+    if position is not None:
+        if not isinstance(position, int) or position < 1:
+            return _validation_error(
+                field="position",
+                action=action,
+                message="position must be a positive integer (1-based)",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+
+    # Validate dry_run
+    dry_run = payload.get("dry_run", False)
+    if dry_run is not None and not isinstance(dry_run, bool):
+        return _validation_error(
+            field="dry_run",
+            action=action,
+            message="dry_run must be a boolean",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    dry_run_bool = bool(dry_run)
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+
+    # Call the core move_task function
+    result, error, warnings = move_task(
+        spec_id=spec_id.strip(),
+        task_id=task_id.strip(),
+        new_parent=new_parent.strip() if new_parent else None,
+        position=position,
+        dry_run=dry_run_bool,
+        specs_dir=specs_dir,
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error or result is None:
+        # Determine appropriate error code based on error message
+        error_lower = (error or "").lower()
+        if "not found" in error_lower:
+            code = ErrorCode.TASK_NOT_FOUND
+            err_type = ErrorType.NOT_FOUND
+            remediation = "Verify the task ID and parent ID exist in the specification"
+        elif "circular" in error_lower:
+            code = ErrorCode.CIRCULAR_DEPENDENCY
+            err_type = ErrorType.CONFLICT
+            remediation = "Task cannot be moved under its own descendants"
+        elif "invalid position" in error_lower:
+            code = ErrorCode.INVALID_POSITION
+            err_type = ErrorType.VALIDATION
+            remediation = "Specify a valid position within the children list"
+        elif "cannot move" in error_lower or "invalid" in error_lower:
+            code = ErrorCode.INVALID_PARENT
+            err_type = ErrorType.VALIDATION
+            remediation = "Specify a valid phase, group, or task as the target parent"
+        else:
+            code = ErrorCode.VALIDATION_ERROR
+            err_type = ErrorType.VALIDATION
+            remediation = "Check task ID, parent, and position parameters"
+
+        return asdict(
+            error_response(
+                error or "Failed to move task",
+                error_code=code,
+                error_type=err_type,
+                remediation=remediation,
+                request_id=request_id,
+                telemetry={"duration_ms": round(elapsed_ms, 2)},
+            )
+        )
+
+    # Build success response with warnings if any
+    response = success_response(
+        **result,
+        request_id=request_id,
+        warnings=warnings if warnings else None,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(
+        _metric(action),
+        labels={"status": "success", "dry_run": str(dry_run_bool).lower()},
+    )
+    return asdict(response)
+
+
+def _handle_add_dependency(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    """Add a dependency relationship between two tasks.
+
+    Manages blocks, blocked_by, and depends relationships.
+    Updates both source and target tasks atomically.
+
+    Dependency types:
+    - blocks: Source task blocks target (target cannot start until source completes)
+    - blocked_by: Source task is blocked by target (source cannot start until target completes)
+    - depends: Soft dependency (informational, doesn't block)
+    """
+    request_id = _request_id()
+    action = "add-dependency"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")  # Source task
+    target_id = payload.get("target_id")  # Target task
+    dependency_type = payload.get("dependency_type", "blocks")
+
+    # Validate required fields
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty source task identifier",
+            request_id=request_id,
+        )
+    if not isinstance(target_id, str) or not target_id.strip():
+        return _validation_error(
+            field="target_id",
+            action=action,
+            message="Provide a non-empty target task identifier",
+            request_id=request_id,
+        )
+
+    # Validate dependency_type
+    valid_types = ("blocks", "blocked_by", "depends")
+    if dependency_type not in valid_types:
+        return _validation_error(
+            field="dependency_type",
+            action=action,
+            message=f"Must be one of: {', '.join(valid_types)}",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    # Validate dry_run
+    dry_run = payload.get("dry_run", False)
+    if dry_run is not None and not isinstance(dry_run, bool):
+        return _validation_error(
+            field="dry_run",
+            action=action,
+            message="dry_run must be a boolean",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    dry_run_bool = bool(dry_run)
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+
+    # Call the core function
+    result, error = manage_task_dependency(
+        spec_id=spec_id.strip(),
+        source_task_id=task_id.strip(),
+        target_task_id=target_id.strip(),
+        dependency_type=dependency_type,
+        action="add",
+        dry_run=dry_run_bool,
+        specs_dir=specs_dir,
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error or result is None:
+        # Determine appropriate error code based on error message
+        error_lower = (error or "").lower()
+        if "not found" in error_lower:
+            code = ErrorCode.TASK_NOT_FOUND
+            err_type = ErrorType.NOT_FOUND
+            remediation = "Verify both task IDs exist in the specification"
+        elif "circular" in error_lower:
+            code = ErrorCode.CIRCULAR_DEPENDENCY
+            err_type = ErrorType.CONFLICT
+            remediation = "This dependency would create a cycle"
+        elif "itself" in error_lower:
+            code = ErrorCode.SELF_REFERENCE
+            err_type = ErrorType.VALIDATION
+            remediation = "A task cannot depend on itself"
+        elif "already exists" in error_lower:
+            code = ErrorCode.DUPLICATE_ENTRY
+            err_type = ErrorType.CONFLICT
+            remediation = "This dependency already exists"
+        else:
+            code = ErrorCode.VALIDATION_ERROR
+            err_type = ErrorType.VALIDATION
+            remediation = "Check task IDs and dependency type"
+
+        return asdict(
+            error_response(
+                error or "Failed to add dependency",
+                error_code=code,
+                error_type=err_type,
+                remediation=remediation,
+                request_id=request_id,
+                telemetry={"duration_ms": round(elapsed_ms, 2)},
+            )
+        )
+
+    # Build success response
+    response = success_response(
+        **result,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(
+        _metric(action),
+        labels={"status": "success", "dry_run": str(dry_run_bool).lower()},
+    )
+    return asdict(response)
+
+
+def _handle_remove_dependency(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    """Remove a dependency relationship between two tasks.
+
+    Removes blocks, blocked_by, or depends relationships.
+    Updates both source and target tasks atomically for reciprocal relationships.
+    """
+    request_id = _request_id()
+    action = "remove-dependency"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")  # Source task
+    target_id = payload.get("target_id")  # Target task
+    dependency_type = payload.get("dependency_type", "blocks")
+
+    # Validate required fields
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty source task identifier",
+            request_id=request_id,
+        )
+    if not isinstance(target_id, str) or not target_id.strip():
+        return _validation_error(
+            field="target_id",
+            action=action,
+            message="Provide a non-empty target task identifier",
+            request_id=request_id,
+        )
+
+    # Validate dependency_type
+    valid_types = ("blocks", "blocked_by", "depends")
+    if dependency_type not in valid_types:
+        return _validation_error(
+            field="dependency_type",
+            action=action,
+            message=f"Must be one of: {', '.join(valid_types)}",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    # Validate dry_run
+    dry_run = payload.get("dry_run", False)
+    if dry_run is not None and not isinstance(dry_run, bool):
+        return _validation_error(
+            field="dry_run",
+            action=action,
+            message="dry_run must be a boolean",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    dry_run_bool = bool(dry_run)
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+
+    # Call the core function
+    result, error = manage_task_dependency(
+        spec_id=spec_id.strip(),
+        source_task_id=task_id.strip(),
+        target_task_id=target_id.strip(),
+        dependency_type=dependency_type,
+        action="remove",
+        dry_run=dry_run_bool,
+        specs_dir=specs_dir,
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error or result is None:
+        # Determine appropriate error code based on error message
+        error_lower = (error or "").lower()
+        if "does not exist" in error_lower:
+            # Dependency relationship doesn't exist
+            code = ErrorCode.DEPENDENCY_NOT_FOUND
+            err_type = ErrorType.NOT_FOUND
+            remediation = "This dependency does not exist"
+        elif "not found" in error_lower:
+            # Task or spec not found
+            code = ErrorCode.TASK_NOT_FOUND
+            err_type = ErrorType.NOT_FOUND
+            remediation = "Verify both task IDs exist in the specification"
+        else:
+            code = ErrorCode.VALIDATION_ERROR
+            err_type = ErrorType.VALIDATION
+            remediation = "Check task IDs and dependency type"
+
+        return asdict(
+            error_response(
+                error or "Failed to remove dependency",
+                error_code=code,
+                error_type=err_type,
+                remediation=remediation,
+                request_id=request_id,
+                telemetry={"duration_ms": round(elapsed_ms, 2)},
+            )
+        )
+
+    # Build success response
+    response = success_response(
+        **result,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(
+        _metric(action),
+        labels={"status": "success", "dry_run": str(dry_run_bool).lower()},
+    )
+    return asdict(response)
+
+
+def _handle_add_requirement(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    """Add a structured requirement to a task's metadata.
+
+    Requirements are stored in metadata.requirements as a list of objects:
+    [{"id": "req-1", "type": "acceptance", "text": "..."}, ...]
+
+    Each requirement has:
+    - id: Auto-generated unique ID (e.g., "req-1", "req-2")
+    - type: Requirement type (acceptance, technical, constraint)
+    - text: Requirement description text
+    """
+    request_id = _request_id()
+    action = "add-requirement"
+    spec_id = payload.get("spec_id")
+    task_id = payload.get("task_id")
+    requirement_type = payload.get("requirement_type")
+    text = payload.get("text")
+
+    # Validate required fields
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    if not isinstance(task_id, str) or not task_id.strip():
+        return _validation_error(
+            field="task_id",
+            action=action,
+            message="Provide a non-empty task identifier",
+            request_id=request_id,
+        )
+    if not isinstance(requirement_type, str) or not requirement_type.strip():
+        return _validation_error(
+            field="requirement_type",
+            action=action,
+            message="Provide a requirement type",
+            request_id=request_id,
+        )
+
+    # Validate requirement_type
+    requirement_type_lower = requirement_type.lower().strip()
+    if requirement_type_lower not in REQUIREMENT_TYPES:
+        return _validation_error(
+            field="requirement_type",
+            action=action,
+            message=f"Must be one of: {', '.join(REQUIREMENT_TYPES)}",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    # Validate text
+    if not isinstance(text, str) or not text.strip():
+        return _validation_error(
+            field="text",
+            action=action,
+            message="Provide non-empty requirement text",
+            request_id=request_id,
+        )
+
+    # Validate dry_run
+    dry_run = payload.get("dry_run", False)
+    if dry_run is not None and not isinstance(dry_run, bool):
+        return _validation_error(
+            field="dry_run",
+            action=action,
+            message="dry_run must be a boolean",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    dry_run_bool = bool(dry_run)
+
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    start = time.perf_counter()
+
+    # Call the core function
+    result, error = update_task_requirements(
+        spec_id=spec_id.strip(),
+        task_id=task_id.strip(),
+        action="add",
+        requirement_type=requirement_type_lower,
+        text=text.strip(),
+        dry_run=dry_run_bool,
+        specs_dir=specs_dir,
+    )
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error or result is None:
+        # Determine appropriate error code based on error message
+        error_lower = (error or "").lower()
+        if "not found" in error_lower:
+            if "specification" in error_lower:
+                code = ErrorCode.SPEC_NOT_FOUND
+                err_type = ErrorType.NOT_FOUND
+                remediation = "Verify the spec ID exists"
+            else:
+                code = ErrorCode.TASK_NOT_FOUND
+                err_type = ErrorType.NOT_FOUND
+                remediation = "Verify the task ID exists in the specification"
+        elif "maximum" in error_lower or "limit" in error_lower:
+            code = ErrorCode.LIMIT_EXCEEDED
+            err_type = ErrorType.VALIDATION
+            remediation = "Remove some requirements before adding new ones"
+        elif "requirement_type" in error_lower:
+            code = ErrorCode.INVALID_FORMAT
+            err_type = ErrorType.VALIDATION
+            remediation = f"Use one of: {', '.join(REQUIREMENT_TYPES)}"
+        else:
+            code = ErrorCode.VALIDATION_ERROR
+            err_type = ErrorType.VALIDATION
+            remediation = "Check task ID and requirement fields"
+
+        return asdict(
+            error_response(
+                error or "Failed to add requirement",
+                error_code=code,
+                error_type=err_type,
+                remediation=remediation,
+                request_id=request_id,
+                telemetry={"duration_ms": round(elapsed_ms, 2)},
+            )
+        )
+
+    # Build success response
+    response = success_response(
+        **result,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(
+        _metric(action),
+        labels={"status": "success", "dry_run": str(dry_run_bool).lower()},
+    )
+    return asdict(response)
+
+
+_VALID_NODE_TYPES = {"task", "verify", "phase", "subtask"}
+# Note: VALID_VERIFICATION_TYPES imported from foundry_mcp.core.validation
+
+
+def _match_nodes_for_batch(
+    hierarchy: Dict[str, Any],
+    *,
+    phase_id: Optional[str] = None,
+    pattern: Optional[str] = None,
+    node_type: Optional[str] = None,
+) -> List[str]:
+    """Filter nodes by phase_id, regex pattern on title/id, and/or node_type.
+
+    All provided filters are combined with AND logic.
+    Returns list of matching node IDs.
+    """
+    matched: List[str] = []
+    compiled_pattern = None
+    if pattern:
+        try:
+            compiled_pattern = re.compile(pattern, re.IGNORECASE)
+        except re.error:
+            return []  # Invalid regex returns empty
+
+    for node_id, node_data in hierarchy.items():
+        if node_id == "spec-root":
+            continue
+
+        # Filter by node_type if specified
+        if node_type and node_data.get("type") != node_type:
+            continue
+
+        # Filter by phase_id if specified (must be under that phase)
+        if phase_id:
+            node_parent = node_data.get("parent")
+            # Direct children of the phase
+            if node_parent != phase_id:
+                # Check if it's a nested child (e.g., subtask under task under phase)
+                parent_node = hierarchy.get(node_parent, {})
+                if parent_node.get("parent") != phase_id:
+                    continue
+
+        # Filter by regex pattern on title or node_id
+        if compiled_pattern:
+            title = node_data.get("title", "")
+            if not (compiled_pattern.search(title) or compiled_pattern.search(node_id)):
+                continue
+
+        matched.append(node_id)
+
+    return sorted(matched)
+
+
+def _handle_metadata_batch(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    """Batch update metadata across multiple tasks matching specified criteria.
+
+    Filters (combined with AND logic):
+    - status_filter: Filter by task status (pending, in_progress, completed, blocked)
+    - parent_filter: Filter by parent node ID (e.g., phase-1, task-2-1)
+    - pattern: Regex pattern to match task titles/IDs
+
+    Legacy filters (deprecated, use parent_filter instead):
+    - phase_id: Alias for parent_filter
+
+    Metadata fields supported:
+    - description, file_path, estimated_hours, category, labels, owners
+    - update_metadata: Dict for custom metadata fields (verification_type, command, etc.)
+    """
+    request_id = _request_id()
+    action = "metadata-batch"
+    start = time.perf_counter()
+
+    # Required: spec_id
+    spec_id = payload.get("spec_id")
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+    spec_id = spec_id.strip()
+
+    # Extract filter parameters
+    status_filter = payload.get("status_filter")
+    parent_filter = payload.get("parent_filter")
+    phase_id = payload.get("phase_id")  # Legacy alias for parent_filter
+    pattern = payload.get("pattern")
+
+    # Use phase_id as parent_filter if parent_filter not provided (backwards compat)
+    if parent_filter is None and phase_id is not None:
+        parent_filter = phase_id
+
+    # Validate status_filter
+    if status_filter is not None:
+        if not isinstance(status_filter, str) or status_filter not in _ALLOWED_STATUS:
+            return _validation_error(
+                field="status_filter",
+                action=action,
+                message=f"status_filter must be one of: {sorted(_ALLOWED_STATUS)}",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+
+    # Validate parent_filter
+    if parent_filter is not None:
+        if not isinstance(parent_filter, str) or not parent_filter.strip():
+            return _validation_error(
+                field="parent_filter",
+                action=action,
+                message="parent_filter must be a non-empty string",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+        parent_filter = parent_filter.strip()
+
+    # Validate pattern
+    if pattern is not None:
+        if not isinstance(pattern, str) or not pattern.strip():
+            return _validation_error(
+                field="pattern",
+                action=action,
+                message="pattern must be a non-empty string",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+        try:
+            re.compile(pattern)
+        except re.error as exc:
+            return _validation_error(
+                field="pattern",
+                action=action,
+                message=f"Invalid regex pattern: {exc}",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+        pattern = pattern.strip()
+
+    # At least one filter must be provided
+    if not any([status_filter, parent_filter, pattern]):
+        return _validation_error(
+            field="status_filter",
+            action=action,
+            message="Provide at least one filter: status_filter, parent_filter, or pattern",
+            request_id=request_id,
+            code=ErrorCode.MISSING_REQUIRED,
+            remediation="Specify status_filter, parent_filter (or phase_id), and/or pattern to target tasks",
+        )
+
+    # Extract metadata fields
+    description = payload.get("description")
+    file_path = payload.get("file_path")
+    estimated_hours = payload.get("estimated_hours")
+    category = payload.get("category")
+    labels = payload.get("labels")
+    owners = payload.get("owners")
+    update_metadata = payload.get("update_metadata")  # Dict for custom fields
+    dry_run = payload.get("dry_run", False)
+
+    # Validate metadata fields
+    if description is not None and not isinstance(description, str):
+        return _validation_error(
+            field="description",
+            action=action,
+            message="description must be a string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    if file_path is not None and not isinstance(file_path, str):
+        return _validation_error(
+            field="file_path",
+            action=action,
+            message="file_path must be a string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    if estimated_hours is not None:
+        if not isinstance(estimated_hours, (int, float)) or estimated_hours < 0:
+            return _validation_error(
+                field="estimated_hours",
+                action=action,
+                message="estimated_hours must be a non-negative number",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+
+    if category is not None and not isinstance(category, str):
+        return _validation_error(
+            field="category",
+            action=action,
+            message="category must be a string",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    if labels is not None:
+        if not isinstance(labels, dict) or not all(
+            isinstance(k, str) and isinstance(v, str) for k, v in labels.items()
+        ):
+            return _validation_error(
+                field="labels",
+                action=action,
+                message="labels must be a dict with string keys and values",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+
+    if owners is not None:
+        if not isinstance(owners, list) or not all(isinstance(o, str) for o in owners):
+            return _validation_error(
+                field="owners",
+                action=action,
+                message="owners must be a list of strings",
+                request_id=request_id,
+                code=ErrorCode.INVALID_FORMAT,
+            )
+
+    if update_metadata is not None and not isinstance(update_metadata, dict):
+        return _validation_error(
+            field="update_metadata",
+            action=action,
+            message="update_metadata must be a dict",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    if dry_run is not None and not isinstance(dry_run, bool):
+        return _validation_error(
+            field="dry_run",
+            action=action,
+            message="dry_run must be a boolean",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+
+    # At least one metadata field must be provided
+    has_metadata = any([
+        description is not None,
+        file_path is not None,
+        estimated_hours is not None,
+        category is not None,
+        labels is not None,
+        owners is not None,
+        update_metadata,
+    ])
+    if not has_metadata:
+        return _validation_error(
+            field="description",
+            action=action,
+            message="Provide at least one metadata field to update",
+            request_id=request_id,
+            code=ErrorCode.MISSING_REQUIRED,
+            remediation="Specify description, file_path, estimated_hours, category, labels, owners, or update_metadata",
+        )
+
+    # Resolve specs directory
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    if specs_dir is None:
+        return _specs_dir_missing_error(request_id)
+
+    # Delegate to core helper
+    result, error = batch_update_tasks(
+        spec_id,
+        status_filter=status_filter,
+        parent_filter=parent_filter,
+        pattern=pattern,
+        description=description,
+        file_path=file_path,
+        estimated_hours=float(estimated_hours) if estimated_hours is not None else None,
+        category=category,
+        labels=labels,
+        owners=owners,
+        custom_metadata=update_metadata,
+        dry_run=bool(dry_run),
+        specs_dir=specs_dir,
+    )
+
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    if error:
+        _metrics.counter(_metric(action), labels={"status": "error"})
+        # Map helper errors to response-v2 format
+        if "not found" in error.lower():
+            return asdict(
+                error_response(
+                    error,
+                    error_code=ErrorCode.NOT_FOUND,
+                    error_type=ErrorType.NOT_FOUND,
+                    remediation="Check spec_id and parent_filter values",
+                    request_id=request_id,
+                )
+            )
+        if "at least one" in error.lower() or "must be" in error.lower():
+            return asdict(
+                error_response(
+                    error,
+                    error_code=ErrorCode.VALIDATION_ERROR,
+                    error_type=ErrorType.VALIDATION,
+                    remediation="Check filter and metadata parameters",
+                    request_id=request_id,
+                )
+            )
+        return asdict(
+            error_response(
+                error,
+                error_code=ErrorCode.INTERNAL_ERROR,
+                error_type=ErrorType.INTERNAL,
+                remediation="Check filesystem permissions and retry",
+                request_id=request_id,
+            )
+        )
+
+    assert result is not None
+
+    # Build response with response-v2 envelope
+    warnings: List[str] = result.get("warnings", [])
+    if result["matched_count"] > _TASK_WARNING_THRESHOLD and not warnings:
+        warnings.append(
+            f"Updated {result['matched_count']} tasks; consider using more specific filters."
+        )
+
+    response = success_response(
+        spec_id=result["spec_id"],
+        matched_count=result["matched_count"],
+        updated_count=result["updated_count"],
+        skipped_count=result.get("skipped_count", 0),
+        nodes=result["nodes"],
+        filters=result["filters"],
+        metadata_applied=result["metadata_applied"],
+        dry_run=result["dry_run"],
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+
+    response_dict = asdict(response)
+    if warnings:
+        meta = response_dict.setdefault("meta", {})
+        meta["warnings"] = warnings
+    if result.get("skipped_tasks"):
+        response_dict["data"]["skipped_tasks"] = result["skipped_tasks"]
+
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return response_dict
+
+
+def _handle_fix_verification_types(
+    *, config: ServerConfig, payload: Dict[str, Any]
+) -> dict:
+    """Fix verification types across all verify nodes in a spec.
+
+    This action:
+    1. Finds all verify nodes with invalid or missing verification_type
+    2. Maps legacy values (e.g., 'test' -> 'run-tests') using VERIFICATION_TYPE_MAPPING
+    3. Sets missing types to 'run-tests' (default)
+    4. Sets unknown types to 'manual' (fallback)
+
+    Supports dry-run mode to preview changes without persisting.
+    """
+    request_id = _request_id()
+    action = "fix-verification-types"
+
+    # Required: spec_id
+    spec_id = payload.get("spec_id")
+    if not isinstance(spec_id, str) or not spec_id.strip():
+        return _validation_error(
+            field="spec_id",
+            action=action,
+            message="Provide a non-empty spec identifier",
+            request_id=request_id,
+        )
+
+    dry_run = payload.get("dry_run", False)
+    if dry_run is not None and not isinstance(dry_run, bool):
+        return _validation_error(
+            field="dry_run",
+            action=action,
+            message="dry_run must be a boolean",
+            request_id=request_id,
+            code=ErrorCode.INVALID_FORMAT,
+        )
+    dry_run_bool = bool(dry_run)
+
+    # Load spec
+    workspace = payload.get("workspace")
+    specs_dir = _resolve_specs_dir(config, workspace)
+    spec_data, error = _load_spec_data(spec_id.strip(), specs_dir, request_id)
+    if error:
+        return error
+    assert spec_data is not None
+
+    start = time.perf_counter()
+    hierarchy = spec_data.get("hierarchy", {})
+
+    # Find verify nodes and collect fixes
+    fixes: List[Dict[str, Any]] = []
+    for node_id, node_data in hierarchy.items():
+        if node_data.get("type") != "verify":
+            continue
+
+        metadata = node_data.get("metadata", {})
+        current_type = metadata.get("verification_type")
+
+        # Determine the fix needed
+        fix_info: Optional[Dict[str, Any]] = None
+
+        if current_type is None:
+            # Missing verification_type -> default to 'run-tests'
+            fix_info = {
+                "node_id": node_id,
+                "title": node_data.get("title", ""),
+                "issue": "missing",
+                "old_value": None,
+                "new_value": "run-tests",
+            }
+        elif current_type not in VALID_VERIFICATION_TYPES:
+            # Invalid type -> check mapping or fallback to 'manual'
+            mapped = VERIFICATION_TYPE_MAPPING.get(current_type)
+            if mapped:
+                fix_info = {
+                    "node_id": node_id,
+                    "title": node_data.get("title", ""),
+                    "issue": "legacy",
+                    "old_value": current_type,
+                    "new_value": mapped,
+                }
+            else:
+                fix_info = {
+                    "node_id": node_id,
+                    "title": node_data.get("title", ""),
+                    "issue": "invalid",
+                    "old_value": current_type,
+                    "new_value": "manual",
+                }
+
+        if fix_info:
+            fixes.append(fix_info)
+
+            if not dry_run_bool:
+                # Apply the fix
+                if "metadata" not in node_data:
+                    node_data["metadata"] = {}
+                node_data["metadata"]["verification_type"] = fix_info["new_value"]
+
+    # Save if not dry_run and there were fixes
+    if not dry_run_bool and fixes:
+        if specs_dir is None or not save_spec(spec_id.strip(), spec_data, specs_dir):
+            return asdict(
+                error_response(
+                    "Failed to save spec after fixing verification types",
+                    error_code=ErrorCode.INTERNAL_ERROR,
+                    error_type=ErrorType.INTERNAL,
+                    remediation="Check filesystem permissions and retry",
+                    request_id=request_id,
+                )
+            )
+
+    elapsed_ms = (time.perf_counter() - start) * 1000
+
+    # Count by issue type
+    missing_count = sum(1 for f in fixes if f["issue"] == "missing")
+    legacy_count = sum(1 for f in fixes if f["issue"] == "legacy")
+    invalid_count = sum(1 for f in fixes if f["issue"] == "invalid")
+
+    response = success_response(
+        spec_id=spec_id.strip(),
+        total_fixes=len(fixes),
+        applied_count=len(fixes) if not dry_run_bool else 0,
+        fixes=fixes,
+        summary={
+            "missing_set_to_run_tests": missing_count,
+            "legacy_mapped": legacy_count,
+            "invalid_set_to_manual": invalid_count,
+        },
+        valid_types=sorted(VALID_VERIFICATION_TYPES),
+        legacy_mappings=VERIFICATION_TYPE_MAPPING,
+        dry_run=dry_run_bool,
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+
+    _metrics.timer(_metric(action) + ".duration_ms", elapsed_ms)
+    _metrics.counter(_metric(action), labels={"status": "success"})
+    return asdict(response)
+
+
+def _handle_session_config(*, config: ServerConfig, payload: Dict[str, Any]) -> dict:
+    """
+    Handle session-config action: get/set autonomous mode preferences.
+
+    This action manages the ephemeral autonomous session state, allowing
+    agents to enable/disable autonomous mode and track task completion
+    during autonomous execution.
+
+    Parameters:
+        get: If true, just return current session config without changes
+        auto_mode: Set autonomous mode enabled (true) or disabled (false)
+
+    Returns:
+        Current session configuration including autonomous state
+    """
+    from datetime import datetime, timezone
+
+    request_id = _request_id()
+    action = "session-config"
+    start = time.perf_counter()
+
+    # Get parameters
+    get_only = payload.get("get", False)
+    auto_mode = payload.get("auto_mode")
+
+    # Get the context tracker and session
+    tracker = get_context_tracker()
+    session = tracker.get_or_create_session()
+
+    # Initialize autonomous if not present
+    if session.autonomous is None:
+        session.autonomous = AutonomousSession()
+
+    # If just getting, return current state
+    if get_only:
+        elapsed_ms = (time.perf_counter() - start) * 1000
+        response = success_response(
+            session_id=session.session_id,
+            autonomous=session.autonomous.to_dict(),
+            message="Current session configuration",
+            request_id=request_id,
+            telemetry={"duration_ms": round(elapsed_ms, 2)},
+        )
+        _metrics.counter(_metric(action), labels={"status": "success", "operation": "get"})
+        return asdict(response)
+
+    # Handle auto_mode setting
+    if auto_mode is not None:
+        if not isinstance(auto_mode, bool):
+            return _validation_error(
+                field="auto_mode",
+                action=action,
+                message="auto_mode must be a boolean (true/false)",
+                request_id=request_id,
+            )
+
+        previous_enabled = session.autonomous.enabled
+        session.autonomous.enabled = auto_mode
+
+        if auto_mode and not previous_enabled:
+            # Starting autonomous mode
+            session.autonomous.started_at = (
+                datetime.now(timezone.utc).isoformat().replace("+00:00", "Z")
+            )
+            session.autonomous.tasks_completed = 0
+            session.autonomous.pause_reason = None
+        elif not auto_mode and previous_enabled:
+            # Stopping autonomous mode
+            session.autonomous.pause_reason = "user"
+
+    elapsed_ms = (time.perf_counter() - start) * 1000
+    response = success_response(
+        session_id=session.session_id,
+        autonomous=session.autonomous.to_dict(),
+        message="Autonomous mode enabled" if session.autonomous.enabled else "Autonomous mode disabled",
+        request_id=request_id,
+        telemetry={"duration_ms": round(elapsed_ms, 2)},
+    )
+    _metrics.counter(_metric(action), labels={"status": "success", "operation": "set"})
+    return asdict(response)
+
+
+_ACTION_DEFINITIONS = [
+    ActionDefinition(
+        name="prepare",
+        handler=_handle_prepare,
+        summary="Prepare next actionable task context",
+    ),
+    ActionDefinition(
+        name="prepare-batch",
+        handler=_handle_prepare_batch,
+        summary="Prepare multiple independent tasks for parallel execution",
+    ),
+    ActionDefinition(
+        name="start-batch",
+        handler=_handle_start_batch,
+        summary="Atomically start multiple tasks as in_progress",
+    ),
+    ActionDefinition(
+        name="complete-batch",
+        handler=_handle_complete_batch,
+        summary="Complete multiple tasks with partial failure support",
+    ),
+    ActionDefinition(
+        name="reset-batch",
+        handler=_handle_reset_batch,
+        summary="Reset stale or specified in_progress tasks to pending",
+    ),
+    ActionDefinition(
+        name="next", handler=_handle_next, summary="Return the next actionable task"
+    ),
+    ActionDefinition(
+        name="info", handler=_handle_info, summary="Fetch task metadata by ID"
+    ),
+    ActionDefinition(
+        name="check-deps",
+        handler=_handle_check_deps,
+        summary="Analyze task dependencies and blockers",
+    ),
+    ActionDefinition(name="start", handler=_handle_start, summary="Start a task"),
+    ActionDefinition(
+        name="complete", handler=_handle_complete, summary="Complete a task"
+    ),
+    ActionDefinition(
+        name="update-status",
+        handler=_handle_update_status,
+        summary="Update task status",
+    ),
+    ActionDefinition(name="block", handler=_handle_block, summary="Block a task"),
+    ActionDefinition(name="unblock", handler=_handle_unblock, summary="Unblock a task"),
+    ActionDefinition(
+        name="list-blocked",
+        handler=_handle_list_blocked,
+        summary="List blocked tasks",
+    ),
+    ActionDefinition(name="add", handler=_handle_add, summary="Add a task"),
+    ActionDefinition(name="remove", handler=_handle_remove, summary="Remove a task"),
+    ActionDefinition(
+        name="move",
+        handler=_handle_move,
+        summary="Move task to new position or parent",
+    ),
+    ActionDefinition(
+        name="add-dependency",
+        handler=_handle_add_dependency,
+        summary="Add a dependency between two tasks",
+    ),
+    ActionDefinition(
+        name="remove-dependency",
+        handler=_handle_remove_dependency,
+        summary="Remove a dependency between two tasks",
+    ),
+    ActionDefinition(
+        name="add-requirement",
+        handler=_handle_add_requirement,
+        summary="Add a structured requirement to a task",
+    ),
+    ActionDefinition(
+        name="update-estimate",
+        handler=_handle_update_estimate,
+        summary="Update estimated effort",
+    ),
+    ActionDefinition(
+        name="update-metadata",
+        handler=_handle_update_metadata,
+        summary="Update task metadata fields",
+    ),
+    ActionDefinition(
+        name="metadata-batch",
+        handler=_handle_metadata_batch,
+        summary="Batch update metadata across multiple nodes matching filters",
+    ),
+    ActionDefinition(
+        name="fix-verification-types",
+        handler=_handle_fix_verification_types,
+        summary="Fix invalid/missing verification types across verify nodes",
+    ),
+    ActionDefinition(
+        name="progress",
+        handler=_handle_progress,
+        summary="Summarize completion metrics for a node",
+    ),
+    ActionDefinition(
+        name="list",
+        handler=_handle_list,
+        summary="List tasks with pagination and optional filters",
+    ),
+    ActionDefinition(
+        name="query",
+        handler=_handle_query,
+        summary="Query tasks by status or parent",
+    ),
+    ActionDefinition(
+        name="hierarchy",
+        handler=_handle_hierarchy,
+        summary="Return paginated hierarchy slices",
+    ),
+    ActionDefinition(
+        name="session-config",
+        handler=_handle_session_config,
+        summary="Get/set autonomous session configuration",
+    ),
+]
+
+_TASK_ROUTER = ActionRouter(tool_name="task", actions=_ACTION_DEFINITIONS)
+
+
+def _dispatch_task_action(
+    *, action: str, payload: Dict[str, Any], config: ServerConfig
+) -> dict:
+    try:
+        return _TASK_ROUTER.dispatch(action=action, config=config, payload=payload)
+    except ActionRouterError as exc:
+        request_id = _request_id()
+        allowed = ", ".join(exc.allowed_actions)
+        return asdict(
+            error_response(
+                f"Unsupported task action '{action}'. Allowed actions: {allowed}",
+                error_code=ErrorCode.VALIDATION_ERROR,
+                error_type=ErrorType.VALIDATION,
+                remediation=f"Use one of: {allowed}",
+                request_id=request_id,
+            )
+        )
+
+
+def register_unified_task_tool(mcp: FastMCP, config: ServerConfig) -> None:
+    """Register the consolidated task tool."""
+
+    @canonical_tool(
+        mcp,
+        canonical_name="task",
+    )
+    @mcp_tool(tool_name="task", emit_metrics=True, audit=True)
+    def task(
+        action: str,
+        spec_id: Optional[str] = None,
+        task_id: Optional[str] = None,
+        task_ids: Optional[List[str]] = None,
+        workspace: Optional[str] = None,
+        status_filter: Optional[str] = None,
+        include_completed: bool = True,
+        node_id: str = "spec-root",
+        include_phases: bool = True,
+        cursor: Optional[str] = None,
+        limit: Optional[int] = None,
+        parent: Optional[str] = None,
+        status: Optional[str] = None,
+        note: Optional[str] = None,
+        completion_note: Optional[str] = None,
+        reason: Optional[str] = None,
+        blocker_type: str = "dependency",
+        ticket: Optional[str] = None,
+        resolution: Optional[str] = None,
+        title: Optional[str] = None,
+        description: Optional[str] = None,
+        acceptance_criteria: Optional[List[str]] = None,
+        task_type: str = "task",
+        estimated_hours: Optional[float] = None,
+        position: Optional[int] = None,
+        cascade: bool = False,
+        complexity: Optional[str] = None,
+        file_path: Optional[str] = None,
+        task_category: Optional[str] = None,
+        actual_hours: Optional[float] = None,
+        status_note: Optional[str] = None,
+        verification_type: Optional[str] = None,
+        command: Optional[str] = None,
+        custom_metadata: Optional[Dict[str, Any]] = None,
+        dry_run: bool = False,
+        max_depth: int = 2,
+        include_metadata: bool = False,
+        # metadata-batch specific parameters
+        phase_id: Optional[str] = None,
+        pattern: Optional[str] = None,
+        node_type: Optional[str] = None,
+        owners: Optional[List[str]] = None,
+        labels: Optional[Dict[str, str]] = None,
+        category: Optional[str] = None,
+        parent_filter: Optional[str] = None,
+        update_metadata: Optional[Dict[str, Any]] = None,
+        # session-config specific parameters
+        get: bool = False,
+        auto_mode: Optional[bool] = None,
+        # complete-batch specific parameters
+        completions: Optional[List[Dict[str, Any]]] = None,
+        # reset-batch specific parameters
+        threshold_hours: Optional[float] = None,
+    ) -> dict:
+        payload = {
+            "spec_id": spec_id,
+            "task_id": task_id,
+            "task_ids": task_ids,
+            "workspace": workspace,
+            "status_filter": status_filter,
+            "include_completed": include_completed,
+            "node_id": node_id,
+            "include_phases": include_phases,
+            "cursor": cursor,
+            "limit": limit,
+            "parent": parent,
+            "status": status,
+            "note": note,
+            "completion_note": completion_note,
+            "reason": reason,
+            "blocker_type": blocker_type,
+            "ticket": ticket,
+            "resolution": resolution,
+            "title": title,
+            "description": description,
+            "acceptance_criteria": acceptance_criteria,
+            "task_type": task_type,
+            "estimated_hours": estimated_hours,
+            "position": position,
+            "cascade": cascade,
+            "complexity": complexity,
+            "file_path": file_path,
+            "task_category": task_category,
+            "actual_hours": actual_hours,
+            "status_note": status_note,
+            "verification_type": verification_type,
+            "command": command,
+            "custom_metadata": custom_metadata,
+            "dry_run": dry_run,
+            "max_depth": max_depth,
+            "include_metadata": include_metadata,
+            # metadata-batch specific
+            "phase_id": phase_id,
+            "pattern": pattern,
+            "node_type": node_type,
+            "owners": owners,
+            "labels": labels,
+            "category": category,
+            "parent_filter": parent_filter,
+            "update_metadata": update_metadata,
+            # session-config specific
+            "get": get,
+            "auto_mode": auto_mode,
+            # complete-batch specific
+            "completions": completions,
+            # reset-batch specific
+            "threshold_hours": threshold_hours,
+        }
+        return _dispatch_task_action(action=action, payload=payload, config=config)
+
+    logger.debug("Registered unified task tool")
+
+
+__all__ = [
+    "register_unified_task_tool",
+]