diff --git a/src/agent_loop.py b/src/agent_loop.py
index 592ebaec1..73bcbdec8 100644
--- a/src/agent_loop.py
+++ b/src/agent_loop.py
@@ -663,7 +663,11 @@ Generate an image. Line 1 = description, line 2 = model name, line 3 = WxH (e.g.
     "manage_session": "- ```manage_session``` — Rename, archive, delete, fork, switch, or `list` chats (the UI calls them 'chats'; 'session' is internal). Line 1 = action (list/switch/rename/archive/unarchive/delete/important/unimportant/truncate/fork), Line 2 = exact chat id from `list_sessions` (or `current` where supported). For delete/archive/truncate, always list first and reuse the exact id; never invent placeholder ids. `switch`/`open` returns a clickable anchor link the user can tap to open the chat — use for \"open my X chat\".",
     "manage_memory": "- ```manage_memory``` — Manage the user's persistent memory (facts about the USER themselves, their preferences, context that persists across chats). Line 1 = action (list/add/edit/delete/search), rest = content. Use when user says 'remember this' about themselves, states identity facts like 'my name is <name>' / 'call me <name>' / 'I live in <place>', or asks about stored memories. DO NOT use for info about another person (their address, phone, email, birthday) — that goes in `manage_contact`. If the user pastes an address/phone with a name and says 'save this for <person>', use `manage_contact add` with the address arg, NOT manage_memory.",
     "manage_skills": "- ```manage_skills``` — Skill registry (SKILL.md format). Args (JSON): {\"action\": \"list|view|view_ref|search|add|edit|patch|publish|delete\", ...}. `list` returns the index of available skills (published + teacher-escalation drafts); `view name=foo` fetches the full SKILL.md; `view_ref name=foo path=...` loads a reference file under the skill directory. For `add`, provide an explicit kebab-case `name` and only report the exact returned name, because storage may normalize or dedupe it. Use this BEFORE doing domain work — there may already be a procedure (published or draft) that prescribes the correct steps. Drafts written by the teacher loop are authoritative guidance even though they're not yet published.",
-    "manage_tasks": "- ```manage_tasks``` — Create and manage scheduled background tasks (recurring AI jobs). Args (JSON): {\"action\": \"list|create|edit|delete|pause|resume|run\", ...}",
+    "manage_tasks": """\
+```manage_tasks
+{"action": "create", "name": "<task name>", "task_type": "llm", "prompt": "<what to do>", "trigger_type": "schedule", "schedule": "daily", "scheduled_time": "09:00"}
+```
+Create and manage scheduled background tasks (recurring AI jobs). action=create needs `name`, `task_type` (llm/research/action), `prompt`, `trigger_type` (schedule/event), and `schedule` (once/daily/weekly/monthly). action=list returns all tasks. action=edit/delete/pause/resume/run needs `task_id`.""",
     "manage_endpoints": "- ```manage_endpoints``` — Add, remove, or configure AI model API endpoints. Args (JSON): {\"action\": \"list|add|delete|enable|disable\", ...}. Use when user wants to add a new AI provider.",
     "manage_mcp": "- ```manage_mcp``` — Manage MCP (Model Context Protocol) tool servers — external tools that extend your capabilities. Args (JSON): {\"action\": \"list|add|delete|reconnect|list_tools\", ...}",
     "manage_webhooks": "- ```manage_webhooks``` — Configure outgoing webhooks (HTTP notifications on events like chat completion). Args (JSON): {\"action\": \"list|add|delete|enable|disable\", ...}",
diff --git a/src/tools/system.py b/src/tools/system.py
index 813d57df2..556a664b1 100644
--- a/src/tools/system.py
+++ b/src/tools/system.py
@@ -280,8 +280,21 @@ async def do_manage_tasks(content: str, owner: Optional[str] = None) -> Dict:
     except ValueError:
         return {"error": "Invalid JSON arguments", "exit_code": 1}
 
+    # Normalize "add" -> "create": models that were not shown the enum sometimes
+    # emit {"action": "add", ...} (same failure class as manage_skills #4013).
+    if args.get("action") == "add":
+        args["action"] = "create"
     if not args.get("action") and any(args.get(k) is not None for k in ("task", "description", "schedule", "time", "day_of_week")):
         args["action"] = "create"
+    # Flatten a nested "task" object: {"action":"create","task":{"name":"foo",...}}
+    # is equivalent to {"action":"create","name":"foo",...} but some models emit
+    # the former. Hoist the inner fields so the rest of the handler sees them flat.
+    task_obj = args.get("task")
+    if isinstance(task_obj, dict):
+        for _k, _v in task_obj.items():
+            if args.get(_k) is None:
+                args[_k] = _v
+        args.pop("task", None)
     if args.get("task") and not args.get("name"):
         args["name"] = args["task"]
     if args.get("task") and not args.get("prompt"):
@@ -331,6 +344,11 @@ async def do_manage_tasks(content: str, owner: Optional[str] = None) -> Dict:
             task_type = args.get("task_type", "llm")
             trigger_type = args.get("trigger_type", "schedule")
 
+            # When the model omits "prompt" but supplies "name", use the name
+            # as the prompt so a bare {"action":"create","name":"foo"} still
+            # produces a runnable task instead of an error.
+            if not args.get("prompt") and args.get("name"):
+                args["prompt"] = args["name"]
             if task_type in ("llm", "research") and not args.get("prompt"):
                 return {"error": "Prompt is required for llm/research tasks", "exit_code": 1}
             if task_type == "action" and not args.get("action_name"):
diff --git a/tests/test_manage_tasks_model_writes.py b/tests/test_manage_tasks_model_writes.py
new file mode 100644
index 000000000..ee64352dc
--- /dev/null
+++ b/tests/test_manage_tasks_model_writes.py
@@ -0,0 +1,120 @@
+"""Regression tests for manage_tasks model-write normalisation (#5757).
+
+Models that were not shown the full tool schema (e.g. when RAG selects a
+different tool set or a local model is used without function-call schemas)
+sometimes emit action="add" instead of action="create", or wrap task fields in
+a nested "task" object rather than providing them at the top level.  Before the
+fix, both patterns fell through to the "unknown action" error branch and the
+task was silently not created.
+
+This test module verifies that both calling patterns produce a successfully
+created task, matching the same-class fix applied to manage_skills (#4013).
+"""
+
+import json
+import tempfile
+
+import pytest
+from sqlalchemy import create_engine
+from sqlalchemy.orm import sessionmaker
+from sqlalchemy.pool import NullPool
+
+from tests.helpers.import_state import clear_fake_database_modules
+
+clear_fake_database_modules()
+
+import core.database as cdb
+from core.database import ScheduledTask
+from src.tools.system import do_manage_tasks
+
+_TMPDB = tempfile.NamedTemporaryFile(suffix=".db", delete=False)
+_ENGINE = create_engine(
+    f"sqlite:///{_TMPDB.name}",
+    connect_args={"check_same_thread": False},
+    poolclass=NullPool,
+)
+cdb.Base.metadata.create_all(_ENGINE)
+_TS = sessionmaker(bind=_ENGINE, autoflush=False, autocommit=False)
+cdb.SessionLocal = _TS
+
+
+def _count_tasks_named(name):
+    db = _TS()
+    try:
+        return db.query(ScheduledTask).filter(ScheduledTask.name == name).count()
+    finally:
+        db.close()
+
+
+@pytest.mark.asyncio
+async def test_action_add_is_treated_as_create():
+    """action='add' must be normalised to 'create' so the task is persisted."""
+    out = await do_manage_tasks(
+        json.dumps({
+            "action": "add",
+            "name": "test task add",
+            "prompt": "summarise the news",
+            "task_type": "llm",
+            "trigger_type": "schedule",
+            "schedule": "daily",
+        }),
+        owner="alice",
+    )
+    assert out["exit_code"] == 0, out
+    assert "Created task" in out["response"]
+    assert _count_tasks_named("test task add") == 1
+
+
+@pytest.mark.asyncio
+async def test_nested_task_object_is_flattened():
+    """{'action':'add','task':{'name':'foo'}} must be treated as create with name=foo."""
+    out = await do_manage_tasks(
+        json.dumps({
+            "action": "add",
+            "task": {
+                "name": "test task 2",
+            },
+        }),
+        owner="alice",
+    )
+    assert out["exit_code"] == 0, out
+    assert "Created task" in out["response"]
+    assert _count_tasks_named("test task 2") == 1
+
+
+@pytest.mark.asyncio
+async def test_nested_task_object_with_prompt():
+    """Nested task object with explicit prompt field is handled correctly."""
+    out = await do_manage_tasks(
+        json.dumps({
+            "action": "create",
+            "task": {
+                "name": "daily summary",
+                "prompt": "summarise my emails",
+            },
+            "trigger_type": "schedule",
+            "schedule": "daily",
+        }),
+        owner="alice",
+    )
+    assert out["exit_code"] == 0, out
+    assert "Created task" in out["response"]
+    assert _count_tasks_named("daily summary") == 1
+
+
+@pytest.mark.asyncio
+async def test_name_used_as_prompt_fallback_when_prompt_absent():
+    """When prompt is absent but name is present, name is used as the prompt."""
+    out = await do_manage_tasks(
+        json.dumps({
+            "action": "create",
+            "name": "my simple task",
+            "task_type": "llm",
+            "trigger_type": "schedule",
+            "schedule": "daily",
+        }),
+        owner="alice",
+    )
+    assert out["exit_code"] == 0, out
+    assert "Created task" in out["response"]
+    assert _count_tasks_named("my simple task") == 1