diff --git a/navi/core/agent.py b/navi/core/agent.py
index ee5911c..14068c6 100644
--- a/navi/core/agent.py
+++ b/navi/core/agent.py
@@ -293,6 +293,7 @@
         briefing: str | None = None,
         custom_system_prompt: str | None = None,
         context_transfer: str | None = None,
+        parent_session_id: str | None = None,
         timeout_seconds: float = 300.0,
     ) -> tuple[str, bool]:
         """
@@ -313,16 +314,19 @@
 
         context_transfer: text from the parent's scratchpad context_transfer section,
             injected as a priming exchange before the task message.
+        parent_session_id: parent chat session id. When provided, sub-agent
+            tool calls run in that session context so session-aware tools resolve
+            filenames against the user's session directory, not an ephemeral id.
         timeout_seconds: wall-clock timeout for the entire sub-agent run.
         """
         import time as _time
         import uuid as _uuid
-        # Give each sub-agent its own scratchpad namespace so parallel or
-        # sequential sub-agents don't clobber each other's working notes.
         from navi.tools.base import current_session_id as _sid_var, current_model as _model_var
         _prev_sid = _sid_var.get(None)
         _prev_model = _model_var.get(None)
-        _sid_var.set(f"subagent_{_uuid.uuid4().hex[:12]}")
+        subagent_run_id = f"subagent_{_uuid.uuid4().hex[:12]}"
+        tool_session_id = parent_session_id or subagent_run_id
+        _sid_var.set(tool_session_id)
 
         profile = self._profiles.get(profile_id)
         _model_var.set(profile.model)
@@ -346,6 +350,14 @@
             sys_parts.append(custom_system_prompt)
         if briefing:
             sys_parts.append(f"## Task context\n\n{briefing}")
+        if parent_session_id:
+            sys_parts.append(
+                "[Parent session context]\n"
+                f"Parent Session ID: {parent_session_id}\n"
+                f"Session files directory: {settings.session_files_dir}/{parent_session_id}/\n"
+                "For files the user should see, write to this exact session directory. "
+                "Do not use or invent a subagent_* directory."
+            )
         if not sys_parts:
             # Fallback if profile has no subagent_system_prompt defined
             sys_parts.append(profile.system_prompt)
diff --git a/navi/profiles/modeler_3d/subagent_system_prompt.txt b/navi/profiles/modeler_3d/subagent_system_prompt.txt
index f4ef89b..04eb57a 100644
--- a/navi/profiles/modeler_3d/subagent_system_prompt.txt
+++ b/navi/profiles/modeler_3d/subagent_system_prompt.txt
@@ -1,22 +1,36 @@
-You are a focused 3D modeling subagent. Complete exactly the delegated task and return only the requested result.
+You are an OpenSCAD transcription worker.
 
-When the task is OpenSCAD authoring or repair:
+Your job is to convert the provided technical specification and design plan into one clean `.scad` file. Do not redesign the object.
 
-- Treat the briefing as the source of truth. Do not infer missing functional dimensions unless the briefing explicitly says to use defaults.
-- Write clean production OpenSCAD to the exact `.scad` path requested in the briefing.
-- Your runtime session id may be an ephemeral `subagent_*` id. Do not derive output paths from it. Use the exact parent session directory and target filename from the briefing.
-- Use only OpenSCAD source and the available tools. Do not use Python/CadQuery/trimesh/raw mesh scripts to generate the model.
-- Keep the source parametric: named dimensions, clear modules, robust boolean overlaps, and no abandoned modules.
-- Do not include prose, self-correction notes, drafts, "let's", "actually", commented-out failed attempts, or conversation transcript inside the `.scad` file.
-- After writing or editing the file, run `scad_lint` on the `.scad` path and fix lint errors you can address locally.
-- Do not publish artifacts. Do not compile/render unless the briefing explicitly asks for it.
-- You may use `scratchpad` to read transferred design context or write a short design note when the briefing is dense. Do not use scratchpad for progress tracking or as a substitute for writing the `.scad` file.
-- After lint succeeds or reports only non-blocking warnings, stop calling tools and return the final response.
+Required workflow:
 
-Final response format for OpenSCAD authoring:
+1. Read the briefing and parent session context.
+2. If the briefing references transferred design context, read `scratchpad` once.
+3. Write the requested `.scad` file with `filesystem`.
+4. Run `scad_lint` on that exact file.
+5. If lint reports errors you can fix directly, edit the same file once and run `scad_lint` again.
+6. Return the final response.
+
+Hard rules:
+
+- Use the exact output path from the parent session context or briefing.
+- The parent session directory is the only valid output directory.
+- Never write outside the parent session directory.
+- Do not derive paths from a `subagent_*` runtime id.
+- Do not change the axis convention from the briefing.
+- If the briefing says `X=length`, the length axis is X. If it says `Z=height`, height is Z.
+- Cylinders, cavities, holes, slots, and recesses must align with the specified axes.
+- Do not change dimensions, tolerances, defaults, or object orientation unless the briefing explicitly instructs it.
+- Do not add extra features that are not in the brief.
+- Do not use scratchpad for progress tracking.
+- Do not compile, render, publish, or share artifacts unless explicitly requested.
+
+If the briefing is insufficient to write correct geometry, return:
+
+BLOCKED: [specific missing detail]
+
+Final response format:
 
 SCAD_PATH: [path]
 LINT: [pass/fail and key messages]
-NOTES: [one concise sentence about implemented parameters or blockers]
-
-If the delegated task is not OpenSCAD authoring, follow the briefing precisely, use tools as needed, and return concise findings with paths or evidence.
+NOTES: [one concise sentence about implemented parameters or blocker]
diff --git a/navi/profiles/modeler_3d/system_prompt.txt b/navi/profiles/modeler_3d/system_prompt.txt
index d9f2ce4..e267fa6 100644
--- a/navi/profiles/modeler_3d/system_prompt.txt
+++ b/navi/profiles/modeler_3d/system_prompt.txt
@@ -114,7 +114,6 @@
 - required source comments contract;
 - instruction to write clean production OpenSCAD only, with no prose, self-correction notes, abandoned modules, or commented-out failed attempts;
 - instruction to call `filesystem write` for the `.scad` file and then `scad_lint`;
-- `max_iterations` set to 40, so the subagent has enough room to reason, write, lint, and repair without premature cutoff;
 - expected final output: the `.scad` path, lint result, and concise notes about implemented parameters.
 
 The subagent must not publish the artifact and should not compile/render unless the main brief explicitly asks for it. The main agent must run final `scad_lint`, `model_3d`, `render_3d`, and `image_view` after the subagent returns.
diff --git a/navi/tools/spawn_agent.py b/navi/tools/spawn_agent.py
index be89415..53463e2 100644
--- a/navi/tools/spawn_agent.py
+++ b/navi/tools/spawn_agent.py
@@ -129,6 +129,7 @@
                 briefing=briefing,
                 custom_system_prompt=custom_system_prompt,
                 context_transfer=context_transfer or None,
+                parent_session_id=parent_sid,
                 timeout_seconds=300.0,
             )
             log.info("spawn_agent.done", profile_id=profile_id, completed=completed,