Merge pull request #83404 from NousResearch/fix/blender-mcp-compromise

fix(security): remove blender MCP catalog entry and skill after upstream compromise
This commit is contained in:
Jeffrey Quesnelle 2026-08-10 15:30:33 -04:00 committed by GitHub
commit 5b8cbd5ef2
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
19 changed files with 14 additions and 1166 deletions

View File

@ -1,88 +0,0 @@
# Nous-approved MCP catalog entry.
# Presence in this directory = approval. Merged via PR review.
manifest_version: 1
name: blender
description: Drive a live Blender session — modeling, scenes, and renders.
source: https://github.com/ahujasid/blender-mcp
# ahujasid/blender-mcp (MIT, PyPI: blender-mcp) is the de-facto standard
# Blender MCP. It is a bridge with two halves:
# 1. This stdio server (uvx blender-mcp), which Hermes launches.
# 2. An addon (addon.py) running inside Blender that opens a local command
# socket on 127.0.0.1:9876. The stdio server relays to it.
# There is nothing to git-clone on the Hermes side — uvx resolves the package —
# so no install block. The in-Blender addon setup is covered in post_install.
#
# Why not Blender's official MCP (blender.org/lab): it requires Blender 5.1+
# and targets scene analysis/documentation rather than asset creation. This
# server works across Blender 3.x/4.x and exposes full modeling control.
transport:
type: stdio
command: "uvx"
# Pinned per catalog dependency policy: exact version, and the release must
# be at least 2 weeks old at pin time (supply-chain cooldown — same rule as
# pyproject.toml dependencies). 1.6.4 released 2026-06-11. The catalog never
# auto-updates; bumping the pin is a PR to this manifest.
args:
- "blender-mcp==1.6.4"
version: "1.6.4"
env:
# Upstream ships anonymous telemetry, on by default. Hermes policy is
# no outbound telemetry without explicit opt-in, so the catalog entry
# disables it. Remove this line only if you deliberately want it on.
DISABLE_TELEMETRY: "true"
# The Blender addon binds 127.0.0.1 only and has no auth of its own; the
# stdio server needs no credentials. Optional integrations (Sketchfab,
# Hyper3D Rodin, Hunyuan3D) take API keys entered in the addon's N-panel
# inside Blender, not via Hermes env.
auth:
type: none
# Tool selection at install time:
# The server advertises 22 tools. 18 of them front optional third-party asset
# services (PolyHaven, Sketchfab, Hyper3D Rodin, Hunyuan3D) that are dead
# weight unless the matching integration is enabled in the addon panel — and
# upstream has no server-side trim mechanism, so without a filter every one
# of them costs schema tokens on every API call. Default to the core surface:
# scene/object inspection, viewport screenshots, and code execution, which is
# the complete modeling/render capability. Users who enable an asset service
# in the addon can opt into its tools with `hermes mcp configure blender`.
tools:
default_enabled:
- get_scene_info
- get_object_info
- get_viewport_screenshot
- execute_blender_code
post_install: |
This entry launches the bridge server, but the bridge talks to an addon
INSIDE a running Blender (3.0+). One-time setup:
1. Download addon.py from https://github.com/ahujasid/blender-mcp
(raw file: https://raw.githubusercontent.com/ahujasid/blender-mcp/main/addon.py)
2. Blender > Edit > Preferences > Add-ons > Install... > select addon.py,
then enable "Interface: Blender MCP".
3. In the 3D viewport press N, open the "BlenderMCP" tab, click
"Connect to Claude" (starts the local socket on 127.0.0.1:9876).
Blender must be RUNNING with the addon connected before the tools work —
start Blender first, then your Hermes session. The addon refuses to start
under `blender -b` (background mode): its command queue runs on UI timers.
On a machine without a display, run Blender under a virtual one:
xvfb-run blender (or Xvfb :99 & DISPLAY=:99 blender)
GPU rendering (Cycles/OptiX) works fine under Xvfb.
SECURITY: execute_blender_code runs arbitrary Python inside Blender with no
sandbox — same trust level as the terminal tool. Upstream telemetry is
disabled via DISABLE_TELEMETRY in this entry's env block.
The 18 asset-service tools (PolyHaven/Sketchfab/Hyper3D/Hunyuan3D) are off
by default. To use one: enable the service in the addon's BlenderMCP panel
(API key there if required), then `hermes mcp configure blender` and tick
its tools. PolyHaven is free and keyless; the others need accounts.
If Hermes and Blender run on different machines, note that file paths in
execute_blender_code (texture loads, exports) resolve on the BLENDER host's
filesystem, not where Hermes runs.

View File

@ -1,142 +0,0 @@
---
name: blender-mcp
description: Drive Blender via the catalog blender MCP, with bpy recipes.
version: 2.1.0
requires: Blender 3.0+ desktop instance (headless via xvfb-run)
author: alireza78a + kshitijk4poor + Hermes Agent
license: MIT
tags: [blender, 3d, animation, modeling, bpy, mcp]
platforms: [linux, macos, windows]
---
# Blender MCP Skill
Companion skill for the `blender` entry in the Hermes MCP catalog. The MCP
server provides the connection to Blender; this skill teaches the bpy idioms
and pitfalls for driving it well. It does not cover Blender UI workflows —
everything here goes through the MCP tools against a live Blender session.
## When to Use
Use when the user wants to create or modify anything in a running Blender
instance: meshes, materials, animations, lighting, renders. Requires the
blender MCP server installed and a Blender desktop session with the addon
connected.
## Prerequisites
1. Install the MCP server from the Nous catalog (one-time):
hermes mcp install blender
This configures the pinned `blender-mcp` stdio server with the curated
tool set: `get_scene_info`, `get_object_info`, `get_viewport_screenshot`,
`execute_blender_code`.
2. Install the addon inside Blender (one-time — the catalog entry's
post-install notes cover this too):
- Download https://raw.githubusercontent.com/ahujasid/blender-mcp/main/addon.py
- Blender > Edit > Preferences > Add-ons > Install... > select addon.py,
enable "Interface: Blender MCP".
3. Every session: start Blender FIRST, press N in the viewport, open the
"BlenderMCP" tab, click "Connect to Claude" (starts the local bridge
socket). Then start your Hermes session so the MCP tools are loaded.
The addon refuses to start under `blender -b` (background mode). On a
machine without a display, run Blender under a virtual one:
`xvfb-run blender`. GPU rendering works fine under Xvfb.
## Quick Reference
| MCP tool | Use for |
|---------------------------|--------------------------------------------|
| `get_scene_info` | List objects before touching the scene |
| `get_object_info` | Inspect one object (transform, materials) |
| `get_viewport_screenshot` | Visual check of what you built |
| `execute_blender_code` | Everything else — arbitrary bpy Python |
Deeper material lives in the reference files (load on demand):
| Reference | Contents |
|-----------|----------|
| `references/bpy-api.md` | Essential bpy operations: modeling, materials, modifiers, rendering |
| `references/recipes.md` | Complete working scenes: low-poly terrain, glass sphere, HDRI lighting, turntable animation |
| `references/pitfalls.md` | Hard-won lessons: empty code results in 5.x, ops-vs-data API, engine names by version |
Optional asset-service tools (PolyHaven, Sketchfab, Hyper3D, Hunyuan3D) are
disabled by default. If the user has enabled a service in the addon panel,
opt into its tools with `hermes mcp configure blender`.
## Procedure
1. Call `get_scene_info` first — never assume the scene is empty.
2. Build with `execute_blender_code`, in small focused calls (one logical
step per call: add objects, then materials, then animation). Large
monolithic scripts hit the bridge timeout.
3. Verify visually with `get_viewport_screenshot` between major steps.
4. Render to an absolute path and tell the user where the file is.
### Common bpy Patterns
Clear scene:
bpy.ops.object.select_all(action='SELECT')
bpy.ops.object.delete()
Add mesh objects:
bpy.ops.mesh.primitive_uv_sphere_add(radius=1, location=(0, 0, 0))
bpy.ops.mesh.primitive_cube_add(size=2, location=(3, 0, 0))
bpy.ops.mesh.primitive_cylinder_add(radius=0.5, depth=2, location=(-3, 0, 0))
Create and assign material:
mat = bpy.data.materials.new(name="MyMat")
mat.use_nodes = True
bsdf = mat.node_tree.nodes.get("Principled BSDF")
bsdf.inputs["Base Color"].default_value = (R, G, B, 1.0)
bsdf.inputs["Roughness"].default_value = 0.3
bsdf.inputs["Metallic"].default_value = 0.0
obj.data.materials.append(mat)
Keyframe animation:
obj.location = (0, 0, 0)
obj.keyframe_insert(data_path="location", frame=1)
obj.location = (0, 0, 3)
obj.keyframe_insert(data_path="location", frame=60)
Render to file:
bpy.context.scene.render.filepath = "/tmp/render.png"
bpy.context.scene.render.engine = 'CYCLES'
bpy.ops.render.render(write_still=True)
## Pitfalls
- The blender MCP tools only exist if the server is installed and the session
started after install. If they're missing, run `hermes mcp install blender`
and start a new session.
- The addon bridge must be (re)connected inside Blender each Blender session
(N-panel > BlenderMCP > Connect). "Connection refused" from the tools means
Blender isn't running or the addon isn't connected — fix that, don't retry.
- Break complex scenes into multiple smaller `execute_blender_code` calls to
avoid bridge timeouts.
- Render output paths must be absolute (`/tmp/render.png`), not relative —
they resolve on the BLENDER host's filesystem, which matters if Hermes and
Blender run on different machines.
- `shade_smooth()` requires the object to be selected and in object mode.
- `execute_blender_code` runs arbitrary Python inside Blender with no sandbox
— same trust level as the `terminal` tool. Don't paste untrusted code into
it.
- Do NOT hand-roll raw TCP JSON to port 9876 from `execute_code` — that was
this skill's pre-MCP workaround. It bypasses the catalog's version pinning
and tool curation. The MCP tools are the supported path.
## Verification
- `get_scene_info` returns the expected object list after each build step.
- `get_viewport_screenshot` shows the scene you intended.
- After a render, confirm the output file exists and report its absolute
path to the user.

View File

@ -1,264 +0,0 @@
# Essential bpy API Reference
## Scene & Objects
```python
# List all objects
[obj.name for obj in bpy.data.objects]
# Get active object
obj = bpy.context.active_object
# Select object by name
bpy.data.objects['Cube'].select_set(True)
# Delete object by name (no context needed)
bpy.data.objects.remove(bpy.data.objects['Cube'], do_unlink=True)
# Delete all objects
for obj in list(bpy.data.objects):
bpy.data.objects.remove(obj, do_unlink=True)
# Duplicate object
import bpy
src = bpy.data.objects['Cube']
new_obj = src.copy()
new_obj.data = src.data.copy()
new_obj.name = 'Cube_Copy'
bpy.context.collection.objects.link(new_obj)
# Parent objects
child = bpy.data.objects['Child']
parent = bpy.data.objects['Parent']
child.parent = parent
```
## Transforms
```python
import math, mathutils
obj = bpy.data.objects['Cube']
# Location (world coordinates)
obj.location = (1.0, 2.0, 3.0)
obj.location.x += 0.5
# Rotation (radians)
obj.rotation_euler = (math.radians(45), 0, 0)
# Scale
obj.scale = (2.0, 2.0, 2.0)
# Apply transforms (bake into mesh)
bpy.context.view_layer.objects.active = obj
bpy.ops.object.transform_apply(location=True, rotation=True, scale=True)
```
## Mesh Creation (bmesh)
```python
import bmesh
# Create mesh from scratch
mesh = bpy.data.meshes.new('CustomMesh')
obj = bpy.data.objects.new('CustomObject', mesh)
bpy.context.collection.objects.link(obj)
bm = bmesh.new()
# Add vertices
v1 = bm.verts.new((0, 0, 0))
v2 = bm.verts.new((1, 0, 0))
v3 = bm.verts.new((0.5, 1, 0))
bm.faces.new((v1, v2, v3))
bm.to_mesh(mesh)
bm.free()
```
## Materials (Principled BSDF)
```python
# Create material
mat = bpy.data.materials.new('MyMat')
mat.use_nodes = True
nodes = mat.node_tree.nodes
links = mat.node_tree.links
principled = nodes['Principled BSDF']
principled.inputs['Base Color'].default_value = (0.8, 0.1, 0.1, 1.0)
principled.inputs['Metallic'].default_value = 0.9
principled.inputs['Roughness'].default_value = 0.1
# Assign to object
obj = bpy.data.objects['Cube']
obj.data.materials.append(mat)
# Emission material
mat = bpy.data.materials.new('GlowMat')
mat.use_nodes = True
nodes = mat.node_tree.nodes
links = mat.node_tree.links
nodes.clear()
output = nodes.new('ShaderNodeOutputMaterial')
emission = nodes.new('ShaderNodeEmission')
emission.inputs['Color'].default_value = (0.0, 1.0, 0.5, 1.0)
emission.inputs['Strength'].default_value = 5.0
links.new(emission.outputs[0], output.inputs[0])
obj.data.materials.append(mat)
# Glass material
mat = bpy.data.materials.new('GlassMat')
mat.use_nodes = True
principled = mat.node_tree.nodes['Principled BSDF']
principled.inputs['Transmission Weight'].default_value = 1.0 # Blender 4.x
principled.inputs['IOR'].default_value = 1.45
principled.inputs['Roughness'].default_value = 0.0
```
## Modifiers
```python
obj = bpy.data.objects['Cube']
# Subdivision Surface
mod = obj.modifiers.new('Subdiv', 'SUBSURF')
mod.levels = 2
mod.render_levels = 3
# Solidify
mod = obj.modifiers.new('Solidify', 'SOLIDIFY')
mod.thickness = 0.05
# Boolean
mod = obj.modifiers.new('Bool', 'BOOLEAN')
mod.operation = 'DIFFERENCE'
mod.object = bpy.data.objects['BoolCutter']
# Array
mod = obj.modifiers.new('Array', 'ARRAY')
mod.count = 5
mod.relative_offset_displace = (1.2, 0, 0)
# Apply modifier
bpy.context.view_layer.objects.active = obj
bpy.ops.object.modifier_apply(modifier='Subdiv')
```
## Camera & Lighting
```python
# Create camera
bpy.ops.object.camera_add(location=(7, -7, 5))
cam = bpy.context.active_object
cam.rotation_euler = (math.radians(63), 0, math.radians(45))
bpy.context.scene.camera = cam
# Camera settings
cam.data.lens = 50 # focal length mm
cam.data.clip_start = 0.1
cam.data.clip_end = 1000
# Point light
bpy.ops.object.light_add(type='POINT', location=(3, 3, 5))
light = bpy.context.active_object
light.data.energy = 1000 # watts
light.data.color = (1.0, 0.9, 0.8)
# Sun light
bpy.ops.object.light_add(type='SUN', rotation=(math.radians(45), 0, 0))
sun = bpy.context.active_object
sun.data.energy = 3
# HDRI world lighting
world = bpy.context.scene.world
if not world:
world = bpy.data.worlds.new('World')
bpy.context.scene.world = world
world.use_nodes = True
nodes = world.node_tree.nodes
links = world.node_tree.links
nodes.clear()
output = nodes.new('ShaderNodeOutputWorld')
bg = nodes.new('ShaderNodeBackground')
env = nodes.new('ShaderNodeTexEnvironment')
env.image = bpy.data.images.load('/path/to/hdri.hdr')
links.new(env.outputs[0], bg.inputs[0])
links.new(bg.outputs[0], output.inputs[0])
```
## Rendering
```python
scene = bpy.context.scene
# Resolution
scene.render.resolution_x = 1920
scene.render.resolution_y = 1080
scene.render.resolution_percentage = 100
# Output
scene.render.filepath = '/tmp/render.png'
scene.render.image_settings.file_format = 'PNG' # PNG, JPEG, OPEN_EXR
# Engine
scene.render.engine = 'CYCLES' # or 'BLENDER_EEVEE' (5.x) / 'BLENDER_EEVEE_NEXT' (4.x)
# Cycles settings
scene.cycles.samples = 128
scene.cycles.use_denoising = True
# Render
bpy.ops.render.render(write_still=True)
# Animation render
scene.frame_start = 1
scene.frame_end = 250
scene.render.filepath = '/tmp/anim_'
scene.render.image_settings.file_format = 'PNG'
bpy.ops.render.render(animation=True)
```
## Animation (Keyframes)
```python
obj = bpy.data.objects['Cube']
scene = bpy.context.scene
# Set keyframe at frame 1
scene.frame_set(1)
obj.location = (0, 0, 0)
obj.keyframe_insert(data_path='location', frame=1)
# Set keyframe at frame 60
scene.frame_set(60)
obj.location = (5, 0, 3)
obj.keyframe_insert(data_path='location', frame=60)
# Rotation keyframe
obj.rotation_euler = (0, 0, math.radians(360))
obj.keyframe_insert(data_path='rotation_euler', frame=60)
# Material keyframe
mat = obj.data.materials[0]
principled = mat.node_tree.nodes['Principled BSDF']
principled.inputs['Base Color'].default_value = (1, 0, 0, 1)
principled.inputs['Base Color'].keyframe_insert(data_path='default_value', frame=1)
principled.inputs['Base Color'].default_value = (0, 0, 1, 1)
principled.inputs['Base Color'].keyframe_insert(data_path='default_value', frame=60)
```
## Collections
```python
# Create collection
col = bpy.data.collections.new('MyCollection')
bpy.context.scene.collection.children.link(col)
# Move object to collection
col.objects.link(obj)
bpy.context.scene.collection.objects.unlink(obj)
# Hide collection
col.hide_viewport = True
col.hide_render = True
```

View File

@ -1,133 +0,0 @@
# Blender MCP — Pitfalls & Lessons Learned
All interaction goes through the blender MCP tools (`hermes mcp install
blender`): `get_scene_info`, `get_object_info`, `get_viewport_screenshot`,
and `execute_blender_code` for arbitrary bpy Python.
## Setup & Connection
### 1. The addon bridge must be started BEFORE the tools work
The Blender MCP addon opens its local bridge socket only when you click
"Connect to Claude" in the BlenderMCP sidebar tab (N-panel). If the MCP
tools error with "connection refused", the addon isn't connected — fix that
in Blender, don't retry the tool.
**Verify the bridge is up:** `lsof -i :9876 -P -n | grep LISTEN`
### 2. Port 9876 is the addon's default — check for conflicts
Other services may already use 9876. If the tools fail but Blender is
running with the addon started, check with lsof. The port is configurable
in the BlenderMCP addon UI panel.
### 3. Addon installation requires user interaction
Blender addon installation requires the GUI: Edit > Preferences > Add-ons >
Install. The agent cannot automate this. Provide the addon.py path and let
the user install it.
## Python Execution (`execute_blender_code`)
### 4. Only bpy and math are in the namespace
The code runs in a namespace with only `bpy` and `math`. If you need os,
json, bmesh, mathutils, etc., import them inside the code:
```python
import bmesh
bm = bmesh.new()
...
```
### 5. Code result is always empty in Blender 5.x
The addon returns `{"executed": true, "result": ""}` for ALL code — the
eval result is not captured in Blender 5.x. To get values out:
- Use `get_scene_info` or `get_object_info` for queries
- Write results to a temp file and read back:
```python
import json
open('/tmp/result.json', 'w').write(json.dumps([o.name for o in bpy.data.objects]))
```
### 6. Errors come back as error strings — always check
The addon catches exceptions and returns them as error text rather than
crashing. Check the tool result for an error before assuming the code ran.
### 7. bpy.ops require correct context
Many `bpy.ops` functions require the right UI context, which differs when
executing through the bridge. Prefer direct data manipulation:
```python
# Prefer data API over ops
bpy.data.objects.remove(bpy.data.objects['Cube'], do_unlink=True)
```
## Objects & Scene
### 8. Default scene has Cube, Light, Camera
New Blender files start with a Cube at (0,0,0), a Light, and a Camera.
Clear them before building.
### 9. Object names are unique — Blender auto-renames duplicates
Creating an object named "Cube" when one already exists results in
"Cube.001". Always check the actual name via `get_scene_info`.
### 10. Degrees vs radians
The addon's own object-creation commands take degrees and convert
internally, but bpy code in `execute_blender_code` uses radians. Be careful
about the distinction.
## Materials
### 11. Principled BSDF is the default shader
Materials created by the addon use Principled BSDF. For other shader types,
build the node tree manually in `execute_blender_code`.
### 12. Color is RGBA 0-1, not RGB 0-255
Material colors use floating-point RGBA in the 0.0-1.0 range.
## Rendering
### 13. Render blocks the bridge — expect long calls
Rendering is synchronous; the tool call won't return until the render
finishes. Expect renders to take far longer than other calls.
### 14. Engine name varies by Blender version
In Blender 5.x, EEVEE is `'BLENDER_EEVEE'` (not `'BLENDER_EEVEE_NEXT'`,
which was Blender 4.x). Discover available engines at runtime:
```python
import json
open('/tmp/engines.json', 'w').write(json.dumps(
list(bpy.types.RenderSettings.bl_rna.properties['engine'].enum_items.keys())))
```
Known engine names: `BLENDER_EEVEE`, `BLENDER_WORKBENCH`, `CYCLES`
### 15. GPU rendering requires explicit setup on macOS
Use METAL compute device type on Apple Silicon. Use CUDA or OPTIX on NVIDIA.
## Reliability
### 16. All state lives in Blender's scene data
Each tool call is independent — there is no session state in the bridge.
Anything you need later must exist in the scene (or a file you wrote).
### 17. Blender crash loses the bridge
If Blender crashes, relaunch it, re-enable the addon, and click "Connect to
Claude" again. Save frequently (`bpy.ops.wm.save_mainfile()`).

View File

@ -1,223 +0,0 @@
# Blender Recipes
Common workflows assembled from the bpy API building blocks.
## Recipe 1: Low-Poly Landscape
```python
import bpy, bmesh, math, random
# Clear scene
for obj in list(bpy.data.objects): bpy.data.objects.remove(obj, do_unlink=True)
# Create subdivided plane
bpy.ops.mesh.primitive_plane_add(size=20)
plane = bpy.context.active_object
plane.name = 'Terrain'
# Subdivide
mod = plane.modifiers.new('Subdiv', 'SUBSURF')
mod.levels = 5
mod.subdivision_type = 'SIMPLE'
bpy.context.view_layer.objects.active = plane
bpy.ops.object.modifier_apply(modifier='Subdiv')
# Displace vertices for terrain
bpy.ops.object.mode_set(mode='EDIT')
bm = bmesh.from_edit_mesh(plane.data)
random.seed(42)
for v in bm.verts:
v.co.z = random.gauss(0, 0.5) * (1 - abs(v.co.x)/10) * (1 - abs(v.co.y)/10)
bmesh.update_edit_mesh(plane.data)
bpy.ops.object.mode_set(mode='OBJECT')
# Green material
mat = bpy.data.materials.new('Grass')
mat.use_nodes = True
mat.node_tree.nodes['Principled BSDF'].inputs['Base Color'].default_value = (0.15, 0.4, 0.1, 1)
mat.node_tree.nodes['Principled BSDF'].inputs['Roughness'].default_value = 0.9
plane.data.materials.append(mat)
# Sun light
bpy.ops.object.light_add(type='SUN', rotation=(math.radians(45), 0, math.radians(30)))
bpy.context.active_object.data.energy = 3
# Camera
bpy.ops.object.camera_add(location=(12, -12, 8))
cam = bpy.context.active_object
cam.rotation_euler = (math.radians(55), 0, math.radians(45))
bpy.context.scene.camera = cam
```
## Recipe 2: Glass Sphere on Reflective Plane
```python
import bpy, math
for obj in list(bpy.data.objects): bpy.data.objects.remove(obj, do_unlink=True)
# Reflective floor
bpy.ops.mesh.primitive_plane_add(size=20)
floor = bpy.context.active_object
mat_floor = bpy.data.materials.new('Floor')
mat_floor.use_nodes = True
p = mat_floor.node_tree.nodes['Principled BSDF']
p.inputs['Base Color'].default_value = (0.02, 0.02, 0.02, 1)
p.inputs['Metallic'].default_value = 1.0
p.inputs['Roughness'].default_value = 0.05
floor.data.materials.append(mat_floor)
# Glass sphere
bpy.ops.mesh.primitive_uv_sphere_add(radius=1.5, location=(0, 0, 1.5), segments=64, ring_count=32)
sphere = bpy.context.active_object
mod = sphere.modifiers.new('Smooth', 'SUBSURF')
mod.levels = 2
mat_glass = bpy.data.materials.new('Glass')
mat_glass.use_nodes = True
p = mat_glass.node_tree.nodes['Principled BSDF']
p.inputs['Transmission Weight'].default_value = 1.0
p.inputs['IOR'].default_value = 1.45
p.inputs['Roughness'].default_value = 0.0
sphere.data.materials.append(mat_glass)
# Three-point lighting
bpy.ops.object.light_add(type='AREA', location=(4, -3, 5))
bpy.context.active_object.data.energy = 500
bpy.context.active_object.data.size = 3
bpy.ops.object.light_add(type='AREA', location=(-4, -2, 3))
bpy.context.active_object.data.energy = 200
bpy.context.active_object.data.size = 2
bpy.ops.object.light_add(type='AREA', location=(0, 4, 2))
bpy.context.active_object.data.energy = 100
bpy.context.active_object.data.size = 4
# Camera
bpy.ops.object.camera_add(location=(5, -5, 3))
cam = bpy.context.active_object
cam.rotation_euler = (math.radians(70), 0, math.radians(45))
bpy.context.scene.camera = cam
# Cycles for glass
bpy.context.scene.render.engine = 'CYCLES'
bpy.context.scene.cycles.samples = 256
```
## Recipe 3: Procedural Donut (Simplified)
```python
import bpy, math
for obj in list(bpy.data.objects): bpy.data.objects.remove(obj, do_unlink=True)
# Torus (donut body)
bpy.ops.mesh.primitive_torus_add(
major_radius=1.0, minor_radius=0.4,
major_segments=48, minor_segments=24
)
donut = bpy.context.active_object
donut.name = 'Donut'
# Subdivision for smoothness
mod = donut.modifiers.new('Subdiv', 'SUBSURF')
mod.levels = 2
# Donut material (warm brown)
mat = bpy.data.materials.new('DonutMat')
mat.use_nodes = True
p = mat.node_tree.nodes['Principled BSDF']
p.inputs['Base Color'].default_value = (0.45, 0.22, 0.08, 1.0)
p.inputs['Roughness'].default_value = 0.7
p.inputs['Subsurface Weight'].default_value = 0.3
donut.data.materials.append(mat)
# Icing (duplicate top half, scale up slightly)
bpy.ops.mesh.primitive_torus_add(
major_radius=1.02, minor_radius=0.42,
major_segments=48, minor_segments=24
)
icing = bpy.context.active_object
icing.name = 'Icing'
# Pink icing material
mat_icing = bpy.data.materials.new('IcingMat')
mat_icing.use_nodes = True
p = mat_icing.node_tree.nodes['Principled BSDF']
p.inputs['Base Color'].default_value = (0.9, 0.4, 0.5, 1.0)
p.inputs['Roughness'].default_value = 0.3
p.inputs['Coat Weight'].default_value = 0.5
icing.data.materials.append(mat_icing)
```
## Recipe 4: Turntable Animation
```python
import bpy, math
# Assume scene already has objects
# Create empty as rotation center
bpy.ops.object.empty_add(location=(0, 0, 0))
pivot = bpy.context.active_object
pivot.name = 'TurntablePivot'
# Parent camera to pivot
cam = bpy.data.objects.get('Camera')
if cam:
cam.parent = pivot
cam.location = (7, 0, 3)
cam.rotation_euler = (math.radians(75), 0, math.radians(90))
# Animate rotation
scene = bpy.context.scene
scene.frame_start = 1
scene.frame_end = 120 # 5 seconds at 24fps
pivot.rotation_euler = (0, 0, 0)
pivot.keyframe_insert(data_path='rotation_euler', frame=1)
pivot.rotation_euler = (0, 0, math.radians(360))
pivot.keyframe_insert(data_path='rotation_euler', frame=121)
# Make rotation linear (not eased)
for fc in pivot.animation_data.action.fcurves:
for kp in fc.keyframe_points:
kp.interpolation = 'LINEAR'
# Render settings
scene.render.filepath = '/tmp/turntable_'
scene.render.image_settings.file_format = 'PNG'
scene.render.resolution_x = 1080
scene.render.resolution_y = 1080
```
## Recipe 5: Render to File and Verify
```python
import bpy, os
scene = bpy.context.scene
scene.render.resolution_x = 1920
scene.render.resolution_y = 1080
scene.render.filepath = '/tmp/blender_render.png'
scene.render.image_settings.file_format = 'PNG'
# Use Cycles for quality
scene.render.engine = 'CYCLES'
scene.cycles.samples = 128
scene.cycles.use_denoising = True
# Render
bpy.ops.render.render(write_still=True)
# Verify
result = os.path.exists('/tmp/blender_render.png')
```
Then from the agent, view the render:
```python
# After execute_blender_code returns, verify and view
from hermes_tools import terminal
terminal("ls -la /tmp/blender_render.png")
# Use vision_analyze to inspect the render
```

View File

@ -8,7 +8,7 @@ platforms: [linux, macos, windows]
metadata:
hermes:
tags: [video, kanban, multi-agent, orchestration, production-pipeline]
related_skills: [ascii-video, manim-video, p5js, comfyui, touchdesigner-mcp, blender-mcp, pixel-art, ascii-art, songwriting-and-ai-music, heartmula, songsee, youtube-content, claude-design, excalidraw, architecture-diagram, concept-diagrams, baoyu-comic, baoyu-infographic, humanizer, gif-search, meme-generation]
related_skills: [ascii-video, manim-video, p5js, comfyui, touchdesigner-mcp, pixel-art, ascii-art, songwriting-and-ai-music, heartmula, songsee, youtube-content, claude-design, excalidraw, architecture-diagram, concept-diagrams, baoyu-comic, baoyu-infographic, humanizer, gif-search, meme-generation]
credits: |
The single-project workspace layout, profile-config patching pattern,
SOUL.md-per-profile model, TEAM.md task-graph convention, and
@ -33,7 +33,7 @@ This skill does **not** render anything itself. It is a meta-pipeline that:
The actual rendering happens inside the kanban once it's running, via whichever
existing skills + tools fit the scenes — `ascii-video`, `manim-video`, `p5js`,
`comfyui`, `touchdesigner-mcp`, `blender-mcp`, `songwriting-and-ai-music`,
`comfyui`, `touchdesigner-mcp`, `songwriting-and-ai-music`,
`heartmula`, external APIs, or plain Python with PIL + ffmpeg.
## When NOT to use this skill

View File

@ -216,7 +216,7 @@ When the user describes a video, look for these signals to map to an example:
- **"Audio-reactive", "real-time", "installation"** → Example 6 (TouchDesigner)
- **Comic-style narrative** → use `renderer-comic` (`baoyu-comic` skill)
- **Retro game / pixel-art aesthetic** → use `renderer-pixel` (`pixel-art` skill)
- **3D scene, photoreal environment** → use `renderer-3d` (`blender-mcp`)
- **3D scene, photoreal environment** → use `renderer-3d` (Blender via `blender -b` scripting)
- **Generative art, particle system, shader** → use `renderer-p5js` (`p5js`)
- **AI-generated photoreal stills + animation** → use `renderer-comfyui`
(`comfyui`) for both stills and image-to-video

View File

@ -117,7 +117,7 @@ instead of overloading one. Each loads a different creative skill.
| `renderer-p5js` | `p5js` | Generative art, particles, shaders, organic motion, web-canvas content |
| `renderer-comfyui` | `comfyui` | AI-generated stills + video using local ComfyUI workflows (img-to-img, img-to-video, etc.) |
| `renderer-touchdesigner` | `touchdesigner-mcp` | Real-time, audio-reactive, installation art, VJ-style content |
| `renderer-3d` | `blender-mcp` *(optional)* | 3D modeling, animation, photoreal environments, character animation |
| `renderer-3d` | Blender (`blender -b` scripting) | 3D modeling, animation, photoreal environments, character animation |
| `renderer-pixel` | `pixel-art` | Retro game aesthetic with era-correct palettes |
| `renderer-comic` | `baoyu-comic` | Knowledge-comic style narrative scenes |
| `renderer-meme` | `meme-generation` *(optional)* | Meme-style stills for satirical/social content |

View File

@ -17,7 +17,6 @@ called from the terminal toolset; they don't appear in `always_load`.
| `p5js` | p5.js sketches — generative art, shaders, interactive, 3D | Renderer for generative art, particle systems, organic motion, web-canvas content |
| `comfyui` | Generate images, video, audio with ComfyUI workflows (image-to-image, image-to-video, etc.) | image-generator, image-to-video-generator, or general renderer for AI-generated content |
| `touchdesigner-mcp` | Control a running TouchDesigner instance — real-time visuals, audio-reactive installation art, VJ | Renderer for real-time/audio-reactive content; installation art; live performance |
| `blender-mcp` *(optional)* | Control Blender 4.3+ via MCP — 3D modeling, animation, rendering | Renderer for 3D scenes, photoreal environments, character animation |
| `pixel-art` | Pixel art with era palettes (NES, Game Boy, PICO-8) | Renderer for retro game aesthetic; concept artist for pixel-style frames |
| `baoyu-comic` | Knowledge-comic generation (educational, biography, tutorial) | Renderer for comic-style narrative; explainer in panel form |
| `baoyu-infographic` | Infographic generation | Renderer for data-driven explainer scenes |
@ -74,7 +73,7 @@ them directly.
| Suno API or web | Track composition (paired with `songwriting-and-ai-music`) | music-supervisor |
| Remotion CLI (`npx remotion render`) | React-based motion graphics | renderer-motion-graphics |
| Manim CE (`manim`) | Math animation render (driven by `manim-video` skill's recipes) | renderer-manim |
| Blender (`blender -b`) | 3D rendering (alternative to `blender-mcp`) | renderer-3d |
| Blender (`blender -b`) | 3D rendering (headless scripting) | renderer-3d |
## Built-in Hermes tools for media review
@ -164,7 +163,6 @@ skills:
# - manim-video (math/explainer)
# - p5js (generative)
# - comfyui (AI-generated visuals)
# - blender-mcp (3D)
# - touchdesigner-mcp (real-time/installation)
```
@ -183,7 +181,6 @@ skills:
# - p5js (renderer-p5js)
# - comfyui (renderer-comfyui — img/video AI gen)
# - touchdesigner-mcp (renderer-touchdesigner)
# - blender-mcp (renderer-3d)
# - pixel-art (renderer-pixel)
# - baoyu-comic (renderer-comic)
# - meme-generation (renderer-meme)

View File

@ -10,7 +10,7 @@ platforms: [linux, macos, windows]
metadata:
hermes:
tags: [unreal, unreal-engine, ue5, 3d, mcp, scenes, cinematics, lighting, gamedev]
related_skills: [blender-mcp]
related_skills: []
---
# Unreal Engine MCP Skill
@ -33,7 +33,7 @@ editor. Works for single actions ("make the sun golden hour") and for
complete multi-step projects ("build me a moody forest clearing with a
campfire and render a shot of it").
Don't use for: DCC-style mesh modeling/sculpting (use `blender-mcp` and
Don't use for: DCC-style mesh modeling/sculpting (model in Blender and
import the result), or for editing Unreal C++ project source (that's normal
code work — use the terminal; this skill is about the live editor).

View File

@ -167,7 +167,7 @@ yes; author new geometry, no. The supported route to parametric geometry is
a custom Python toolset wrapping **Geometry Script** (`UDynamicMesh`:
append box/cylinder/sphere, booleans, then `Create New Static Mesh Asset
from Mesh` to bake an `SM_` asset). For organic/sculpted meshes, model in
Blender (`blender-mcp` skill) and import.
Blender and import.
First-session move: `list_toolsets`, then `describe_toolset` each group you
plan to use, and keep those schemas in working memory for the session.

View File

@ -58,7 +58,6 @@ hermes skills uninstall <skill-name>
| [**audiocraft-audio-generation**](/docs/user-guide/skills/optional/creative/creative-audiocraft-audio-generation) | AudioCraft: MusicGen text-to-music, AudioGen text-to-sound. |
| [**baoyu-article-illustrator**](/docs/user-guide/skills/optional/creative/creative-baoyu-article-illustrator) | Article illustrations: type × style × palette consistency. |
| [**baoyu-comic**](/docs/user-guide/skills/optional/creative/creative-baoyu-comic) | Knowledge comics (知识漫画): educational, biography, tutorial. |
| [**blender-mcp**](/docs/user-guide/skills/optional/creative/creative-blender-mcp) | Drive Blender via the catalog blender MCP, with bpy recipes. |
| [**concept-diagrams**](/docs/user-guide/skills/optional/creative/creative-concept-diagrams) | Generate flat, minimal educational SVG visuals as HTML. |
| [**creative-ideation**](/docs/user-guide/skills/optional/creative/creative-creative-ideation) | Generate ideas via named methods from creative practice. |
| [**heartmula**](/docs/user-guide/skills/optional/creative/creative-heartmula) | HeartMuLa: Suno-like song generation from lyrics + tags. |

View File

@ -1,160 +0,0 @@
---
title: "Blender Mcp — Drive Blender via the catalog blender MCP, with bpy recipes"
sidebar_label: "Blender Mcp"
description: "Drive Blender via the catalog blender MCP, with bpy recipes"
---
{/* This page is auto-generated from the skill's SKILL.md by website/scripts/generate-skill-docs.py. Edit the source SKILL.md, not this page. */}
# Blender Mcp
Drive Blender via the catalog blender MCP, with bpy recipes.
## Skill metadata
| | |
|---|---|
| Source | Optional — install with `hermes skills install official/creative/blender-mcp` |
| Path | `optional-skills/creative/blender-mcp` |
| Version | `2.1.0` |
| Author | alireza78a + kshitijk4poor + Hermes Agent |
| License | MIT |
| Platforms | linux, macos, windows |
## Reference: full SKILL.md
:::info
The following is the complete skill definition that Hermes loads when this skill is triggered. This is what the agent sees as instructions when the skill is active.
:::
# Blender MCP Skill
Companion skill for the `blender` entry in the Hermes MCP catalog. The MCP
server provides the connection to Blender; this skill teaches the bpy idioms
and pitfalls for driving it well. It does not cover Blender UI workflows —
everything here goes through the MCP tools against a live Blender session.
## When to Use
Use when the user wants to create or modify anything in a running Blender
instance: meshes, materials, animations, lighting, renders. Requires the
blender MCP server installed and a Blender desktop session with the addon
connected.
## Prerequisites
1. Install the MCP server from the Nous catalog (one-time):
hermes mcp install blender
This configures the pinned `blender-mcp` stdio server with the curated
tool set: `get_scene_info`, `get_object_info`, `get_viewport_screenshot`,
`execute_blender_code`.
2. Install the addon inside Blender (one-time — the catalog entry's
post-install notes cover this too):
- Download https://raw.githubusercontent.com/ahujasid/blender-mcp/main/addon.py
- Blender > Edit > Preferences > Add-ons > Install... > select addon.py,
enable "Interface: Blender MCP".
3. Every session: start Blender FIRST, press N in the viewport, open the
"BlenderMCP" tab, click "Connect to Claude" (starts the local bridge
socket). Then start your Hermes session so the MCP tools are loaded.
The addon refuses to start under `blender -b` (background mode). On a
machine without a display, run Blender under a virtual one:
`xvfb-run blender`. GPU rendering works fine under Xvfb.
## Quick Reference
| MCP tool | Use for |
|---------------------------|--------------------------------------------|
| `get_scene_info` | List objects before touching the scene |
| `get_object_info` | Inspect one object (transform, materials) |
| `get_viewport_screenshot` | Visual check of what you built |
| `execute_blender_code` | Everything else — arbitrary bpy Python |
Deeper material lives in the reference files (load on demand):
| Reference | Contents |
|-----------|----------|
| `references/bpy-api.md` | Essential bpy operations: modeling, materials, modifiers, rendering |
| `references/recipes.md` | Complete working scenes: low-poly terrain, glass sphere, HDRI lighting, turntable animation |
| `references/pitfalls.md` | Hard-won lessons: empty code results in 5.x, ops-vs-data API, engine names by version |
Optional asset-service tools (PolyHaven, Sketchfab, Hyper3D, Hunyuan3D) are
disabled by default. If the user has enabled a service in the addon panel,
opt into its tools with `hermes mcp configure blender`.
## Procedure
1. Call `get_scene_info` first — never assume the scene is empty.
2. Build with `execute_blender_code`, in small focused calls (one logical
step per call: add objects, then materials, then animation). Large
monolithic scripts hit the bridge timeout.
3. Verify visually with `get_viewport_screenshot` between major steps.
4. Render to an absolute path and tell the user where the file is.
### Common bpy Patterns
Clear scene:
bpy.ops.object.select_all(action='SELECT')
bpy.ops.object.delete()
Add mesh objects:
bpy.ops.mesh.primitive_uv_sphere_add(radius=1, location=(0, 0, 0))
bpy.ops.mesh.primitive_cube_add(size=2, location=(3, 0, 0))
bpy.ops.mesh.primitive_cylinder_add(radius=0.5, depth=2, location=(-3, 0, 0))
Create and assign material:
mat = bpy.data.materials.new(name="MyMat")
mat.use_nodes = True
bsdf = mat.node_tree.nodes.get("Principled BSDF")
bsdf.inputs["Base Color"].default_value = (R, G, B, 1.0)
bsdf.inputs["Roughness"].default_value = 0.3
bsdf.inputs["Metallic"].default_value = 0.0
obj.data.materials.append(mat)
Keyframe animation:
obj.location = (0, 0, 0)
obj.keyframe_insert(data_path="location", frame=1)
obj.location = (0, 0, 3)
obj.keyframe_insert(data_path="location", frame=60)
Render to file:
bpy.context.scene.render.filepath = "/tmp/render.png"
bpy.context.scene.render.engine = 'CYCLES'
bpy.ops.render.render(write_still=True)
## Pitfalls
- The blender MCP tools only exist if the server is installed and the session
started after install. If they're missing, run `hermes mcp install blender`
and start a new session.
- The addon bridge must be (re)connected inside Blender each Blender session
(N-panel > BlenderMCP > Connect). "Connection refused" from the tools means
Blender isn't running or the addon isn't connected — fix that, don't retry.
- Break complex scenes into multiple smaller `execute_blender_code` calls to
avoid bridge timeouts.
- Render output paths must be absolute (`/tmp/render.png`), not relative —
they resolve on the BLENDER host's filesystem, which matters if Hermes and
Blender run on different machines.
- `shade_smooth()` requires the object to be selected and in object mode.
- `execute_blender_code` runs arbitrary Python inside Blender with no sandbox
— same trust level as the `terminal` tool. Don't paste untrusted code into
it.
- Do NOT hand-roll raw TCP JSON to port 9876 from `execute_code` — that was
this skill's pre-MCP workaround. It bypasses the catalog's version pinning
and tool curation. The MCP tools are the supported path.
## Verification
- `get_scene_info` returns the expected object list after each build step.
- `get_viewport_screenshot` shows the scene you intended.
- After a render, confirm the output file exists and report its absolute
path to the user.

View File

@ -21,7 +21,7 @@ Plan and run multi-agent video production pipelines.
| License | MIT |
| Platforms | linux, macos, windows |
| Tags | `video`, `kanban`, `multi-agent`, `orchestration`, `production-pipeline` |
| Related skills | [`ascii-video`](/docs/user-guide/skills/bundled/creative/creative-ascii-video), [`manim-video`](/docs/user-guide/skills/bundled/creative/creative-manim-video), [`p5js`](/docs/user-guide/skills/bundled/creative/creative-p5js), [`comfyui`](/docs/user-guide/skills/bundled/creative/creative-comfyui), [`touchdesigner-mcp`](/docs/user-guide/skills/bundled/creative/creative-touchdesigner-mcp), [`blender-mcp`](/docs/user-guide/skills/optional/creative/creative-blender-mcp), [`pixel-art`](/docs/user-guide/skills/optional/creative/creative-pixel-art), [`ascii-art`](/docs/user-guide/skills/bundled/creative/creative-ascii-art), [`songwriting-and-ai-music`](/docs/user-guide/skills/bundled/creative/creative-songwriting-and-ai-music), [`heartmula`](/docs/user-guide/skills/optional/creative/creative-heartmula), [`songsee`](/docs/user-guide/skills/bundled/media/media-songsee), [`youtube-content`](/docs/user-guide/skills/bundled/media/media-youtube-content), [`claude-design`](/docs/user-guide/skills/bundled/creative/creative-claude-design), [`excalidraw`](/docs/user-guide/skills/bundled/creative/creative-excalidraw), [`architecture-diagram`](/docs/user-guide/skills/bundled/creative/creative-architecture-diagram), [`concept-diagrams`](/docs/user-guide/skills/optional/creative/creative-concept-diagrams), [`baoyu-comic`](/docs/user-guide/skills/optional/creative/creative-baoyu-comic), [`baoyu-infographic`](/docs/user-guide/skills/bundled/creative/creative-baoyu-infographic), [`humanizer`](/docs/user-guide/skills/bundled/creative/creative-humanizer), [`gif-search`](/docs/user-guide/skills/bundled/media/media-gif-search), [`meme-generation`](/docs/user-guide/skills/optional/creative/creative-meme-generation) |
| Related skills | [`ascii-video`](/docs/user-guide/skills/bundled/creative/creative-ascii-video), [`manim-video`](/docs/user-guide/skills/bundled/creative/creative-manim-video), [`p5js`](/docs/user-guide/skills/bundled/creative/creative-p5js), [`comfyui`](/docs/user-guide/skills/bundled/creative/creative-comfyui), [`touchdesigner-mcp`](/docs/user-guide/skills/bundled/creative/creative-touchdesigner-mcp), [`pixel-art`](/docs/user-guide/skills/optional/creative/creative-pixel-art), [`ascii-art`](/docs/user-guide/skills/bundled/creative/creative-ascii-art), [`songwriting-and-ai-music`](/docs/user-guide/skills/bundled/creative/creative-songwriting-and-ai-music), [`heartmula`](/docs/user-guide/skills/optional/creative/creative-heartmula), [`songsee`](/docs/user-guide/skills/bundled/media/media-songsee), [`youtube-content`](/docs/user-guide/skills/bundled/media/media-youtube-content), [`claude-design`](/docs/user-guide/skills/bundled/creative/creative-claude-design), [`excalidraw`](/docs/user-guide/skills/bundled/creative/creative-excalidraw), [`architecture-diagram`](/docs/user-guide/skills/bundled/creative/creative-architecture-diagram), [`concept-diagrams`](/docs/user-guide/skills/optional/creative/creative-concept-diagrams), [`baoyu-comic`](/docs/user-guide/skills/optional/creative/creative-baoyu-comic), [`baoyu-infographic`](/docs/user-guide/skills/bundled/creative/creative-baoyu-infographic), [`humanizer`](/docs/user-guide/skills/bundled/creative/creative-humanizer), [`gif-search`](/docs/user-guide/skills/bundled/media/media-gif-search), [`meme-generation`](/docs/user-guide/skills/optional/creative/creative-meme-generation) |
## Reference: full SKILL.md
@ -45,7 +45,7 @@ This skill does **not** render anything itself. It is a meta-pipeline that:
The actual rendering happens inside the kanban once it's running, via whichever
existing skills + tools fit the scenes — `ascii-video`, `manim-video`, `p5js`,
`comfyui`, `touchdesigner-mcp`, `blender-mcp`, `songwriting-and-ai-music`,
`comfyui`, `touchdesigner-mcp`, `songwriting-and-ai-music`,
`heartmula`, external APIs, or plain Python with PIL + ffmpeg.
## When NOT to use this skill

View File

@ -21,7 +21,6 @@ Automate Unreal Engine editor scenes, actors, and renders.
| License | MIT |
| Platforms | linux, macos, windows |
| Tags | `unreal`, `unreal-engine`, `ue5`, `3d`, `mcp`, `scenes`, `cinematics`, `lighting`, `gamedev` |
| Related skills | [`blender-mcp`](/docs/user-guide/skills/optional/creative/creative-blender-mcp) |
## Reference: full SKILL.md
@ -49,7 +48,7 @@ editor. Works for single actions ("make the sun golden hour") and for
complete multi-step projects ("build me a moody forest clearing with a
campfire and render a shot of it").
Don't use for: DCC-style mesh modeling/sculpting (use `blender-mcp` and
Don't use for: DCC-style mesh modeling/sculpting (model in Blender and
import the result), or for editing Unreal C++ project source (that's normal
code work — use the terminal; this skill is about the live editor).

View File

@ -53,7 +53,6 @@ hermes skills uninstall <skill-name>
| 技能 | 描述 |
|-------|-------------|
| [**audiocraft-audio-generation**](/user-guide/skills/optional/creative/creative-audiocraft-audio-generation) | AudioCraftMusicGen 文本转音乐、AudioGen 文本转音效。 |
| [**blender-mcp**](/user-guide/skills/optional/creative/creative-blender-mcp) | 通过 socket 连接 blender-mcp 插件,直接从 Hermes 控制 Blender。创建 3D 对象、材质、动画,并运行任意 Blender Pythonbpy代码。适用于用户希望在 Blender 中创建或修改任何内容的场景。 |
| [**concept-diagrams**](/user-guide/skills/optional/creative/creative-concept-diagrams) | 生成扁平、极简、支持亮色/暗色模式的 SVG 图表,输出为独立 HTML 文件,采用统一的教育视觉语言,包含 9 种语义色阶、句首大写排版及自动暗色模式。最适合教育和说明类内容。 |
| [**heartmula**](/user-guide/skills/optional/creative/creative-heartmula) | HeartMuLa根据歌词 + 标签生成类 Suno 风格的歌曲。 |
| [**hyperframes**](/user-guide/skills/optional/creative/creative-hyperframes) | 使用 HyperFrames 创建基于 HTML 的视频合成、动态标题卡、社交叠层、字幕访谈视频、音频响应视觉效果及着色器转场。HTML 是视频的唯一来源。适用于用户希望制作任何视频内容的场景。 |

View File

@ -1,135 +0,0 @@
---
title: "Blender Mcp — 通过 socket 连接 blender-mcp 插件,直接从 Hermes 控制 Blender"
sidebar_label: "Blender Mcp"
description: "通过 socket 连接 blender-mcp 插件,直接从 Hermes 控制 Blender"
---
{/* This page is auto-generated from the skill's SKILL.md by website/scripts/generate-skill-docs.py. Edit the source SKILL.md, not this page. */}
# Blender Mcp
通过 socket 连接 blender-mcp 插件,直接从 Hermes 控制 Blender。可创建 3D 对象、材质、动画,并运行任意 Blender Pythonbpy代码。当用户需要在 Blender 中创建或修改任何内容时使用。
## Skill 元数据
| | |
|---|---|
| 来源 | 可选 — 通过 `hermes skills install official/creative/blender-mcp` 安装 |
| 路径 | `optional-skills/creative/blender-mcp` |
| 版本 | `1.0.0` |
| 作者 | alireza78a |
| 平台 | linux, macos, windows |
## 参考:完整 SKILL.md
:::info
以下是 Hermes 在触发此 skill 时加载的完整 skill 定义。这是 agent 在 skill 激活时所看到的指令内容。
:::
# Blender MCP
通过 TCP 端口 9876 上的 socket从 Hermes 控制正在运行的 Blender 实例。
## 设置(一次性)
### 1. 安装 Blender 插件
curl -sL https://raw.githubusercontent.com/ahujasid/blender-mcp/main/addon.py -o ~/Desktop/blender_mcp_addon.py
在 Blender 中:
Edit > Preferences > Add-ons > Install > 选择 blender_mcp_addon.py
启用 "Interface: Blender MCP"
### 2. 在 Blender 中启动 socket 服务器
在 Blender 视口中按 N 键打开侧边栏。
找到 "BlenderMCP" 标签页,点击 "Start Server"。
### 3. 验证连接
nc -z -w2 localhost 9876 && echo "OPEN" || echo "CLOSED"
## 协议
通过 TCP 传输纯 UTF-8 JSON — 无长度前缀。
发送: &#123;"type": "&lt;command>", "params": &#123;&lt;kwargs>&#125;&#125;
接收: &#123;"status": "success", "result": &lt;value>&#125;
&#123;"status": "error", "message": "&lt;reason>"&#125;
## 可用命令
| type | params | 说明 |
|-------------------------|-------------------|---------------------------------|
| execute_code | code (str) | 运行任意 bpy Python 代码 |
| get_scene_info | (无) | 列出场景中的所有对象 |
| get_object_info | object_name (str) | 获取特定对象的详细信息 |
| get_viewport_screenshot | (无) | 截取当前视口截图 |
## Python 辅助函数
在 execute_code 工具调用中使用:
import socket, json
def blender_exec(code: str, host="localhost", port=9876, timeout=15):
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.connect((host, port))
s.settimeout(timeout)
payload = json.dumps(&#123;"type": "execute_code", "params": &#123;"code": code&#125;&#125;)
s.sendall(payload.encode("utf-8"))
buf = b""
while True:
try:
chunk = s.recv(4096)
if not chunk:
break
buf += chunk
try:
json.loads(buf.decode("utf-8"))
break
except json.JSONDecodeError:
continue
except socket.timeout:
break
s.close()
return json.loads(buf.decode("utf-8"))
## 常用 bpy 模式
### 清空场景
bpy.ops.object.select_all(action='SELECT')
bpy.ops.object.delete()
### 添加网格对象
bpy.ops.mesh.primitive_uv_sphere_add(radius=1, location=(0, 0, 0))
bpy.ops.mesh.primitive_cube_add(size=2, location=(3, 0, 0))
bpy.ops.mesh.primitive_cylinder_add(radius=0.5, depth=2, location=(-3, 0, 0))
### 创建并指定材质
mat = bpy.data.materials.new(name="MyMat")
mat.use_nodes = True
bsdf = mat.node_tree.nodes.get("Principled BSDF")
bsdf.inputs["Base Color"].default_value = (R, G, B, 1.0)
bsdf.inputs["Roughness"].default_value = 0.3
bsdf.inputs["Metallic"].default_value = 0.0
obj.data.materials.append(mat)
### 关键帧动画
obj.location = (0, 0, 0)
obj.keyframe_insert(data_path="location", frame=1)
obj.location = (0, 0, 3)
obj.keyframe_insert(data_path="location", frame=60)
### 渲染到文件
bpy.context.scene.render.filepath = "/tmp/render.png"
bpy.context.scene.render.engine = 'CYCLES'
bpy.ops.render.render(write_still=True)
## 注意事项
- 运行前必须检查 socket 是否已开放nc -z localhost 9876
- 每次会话都需要在 Blender 内部启动插件服务器N 面板 > BlenderMCP > Connect
- 将复杂场景拆分为多个较小的 execute_code 调用,以避免超时
- 渲染输出路径必须为绝对路径(/tmp/...),不能使用相对路径
- `shade_smooth()` 要求对象已被选中且处于对象模式

View File

@ -8,7 +8,7 @@ description: "规划、搭建并监控由 Hermes Kanban 支撑的多智能体视
# Kanban Video Orchestrator
规划、搭建并监控由 Hermes Kanban 支撑的多智能体视频制作流水线。当用户想要制作**任何**类型的视频时使用本技能——叙事短片、产品/营销视频、MV、解说视频、ASCII/终端艺术、抽象/生成循环、漫画、3D、实时/装置艺术——且工作需要分解为专业角色(编剧、设计师、动画师、渲染师、配音、剪辑等)并通过 kanban 看板协调。执行自适应探索以明确需求范围,为所请求的风格设计合适的团队,生成用于创建 Hermes profiles 和初始 kanban 任务的安装脚本,然后协助监控执行过程并在任务卡住或失败时介入。将场景路由到适合每个节拍的 Hermes 渲染/音频/设计技能(`ascii-video`、`manim-video`、`p5js`、`comfyui`、`touchdesigner-mcp`、`blender-mcp`、`pixel-art`、`baoyu-comic`、`claude-design`、`excalidraw`、`songsee`、`heartmula`……)以及用于 TTS、图像生成和图像转视频的外部 API。
规划、搭建并监控由 Hermes Kanban 支撑的多智能体视频制作流水线。当用户想要制作**任何**类型的视频时使用本技能——叙事短片、产品/营销视频、MV、解说视频、ASCII/终端艺术、抽象/生成循环、漫画、3D、实时/装置艺术——且工作需要分解为专业角色(编剧、设计师、动画师、渲染师、配音、剪辑等)并通过 kanban 看板协调。执行自适应探索以明确需求范围,为所请求的风格设计合适的团队,生成用于创建 Hermes profiles 和初始 kanban 任务的安装脚本,然后协助监控执行过程并在任务卡住或失败时介入。将场景路由到适合每个节拍的 Hermes 渲染/音频/设计技能(`ascii-video`、`manim-video`、`p5js`、`comfyui`、`touchdesigner-mcp`、`pixel-art`、`baoyu-comic`、`claude-design`、`excalidraw`、`songsee`、`heartmula`……)以及用于 TTS、图像生成和图像转视频的外部 API。
## 技能元数据
@ -21,7 +21,7 @@ description: "规划、搭建并监控由 Hermes Kanban 支撑的多智能体视
| 许可证 | MIT |
| 平台 | linux, macos, windows |
| 标签 | `video`, `kanban`, `multi-agent`, `orchestration`, `production-pipeline` |
| 相关技能 | [`ascii-video`](/user-guide/skills/bundled/creative/creative-ascii-video)、[`manim-video`](/user-guide/skills/bundled/creative/creative-manim-video)、[`p5js`](/user-guide/skills/bundled/creative/creative-p5js)、[`comfyui`](/user-guide/skills/bundled/creative/creative-comfyui)、[`touchdesigner-mcp`](/user-guide/skills/bundled/creative/creative-touchdesigner-mcp)、[`blender-mcp`](/user-guide/skills/optional/creative/creative-blender-mcp)、[`pixel-art`](/user-guide/skills/bundled/creative/creative-pixel-art)、[`ascii-art`](/user-guide/skills/bundled/creative/creative-ascii-art)、[`songwriting-and-ai-music`](/user-guide/skills/bundled/creative/creative-songwriting-and-ai-music)、[`heartmula`](/user-guide/skills/optional/creative/creative-heartmula)、[`songsee`](/user-guide/skills/bundled/media/media-songsee)、[`youtube-content`](/user-guide/skills/bundled/media/media-youtube-content)、[`claude-design`](/user-guide/skills/bundled/creative/creative-claude-design)、[`excalidraw`](/user-guide/skills/bundled/creative/creative-excalidraw)、[`architecture-diagram`](/user-guide/skills/bundled/creative/creative-architecture-diagram)、[`concept-diagrams`](/user-guide/skills/optional/creative/creative-concept-diagrams)、[`baoyu-comic`](/user-guide/skills/bundled/creative/creative-baoyu-comic)、[`baoyu-infographic`](/user-guide/skills/bundled/creative/creative-baoyu-infographic)、[`humanizer`](/user-guide/skills/bundled/creative/creative-humanizer)、[`gif-search`](/user-guide/skills/bundled/media/media-gif-search)、[`meme-generation`](/user-guide/skills/optional/creative/creative-meme-generation) |
| 相关技能 | [`ascii-video`](/user-guide/skills/bundled/creative/creative-ascii-video)、[`manim-video`](/user-guide/skills/bundled/creative/creative-manim-video)、[`p5js`](/user-guide/skills/bundled/creative/creative-p5js)、[`comfyui`](/user-guide/skills/bundled/creative/creative-comfyui)、[`touchdesigner-mcp`](/user-guide/skills/bundled/creative/creative-touchdesigner-mcp)、[`pixel-art`](/user-guide/skills/bundled/creative/creative-pixel-art)、[`ascii-art`](/user-guide/skills/bundled/creative/creative-ascii-art)、[`songwriting-and-ai-music`](/user-guide/skills/bundled/creative/creative-songwriting-and-ai-music)、[`heartmula`](/user-guide/skills/optional/creative/creative-heartmula)、[`songsee`](/user-guide/skills/bundled/media/media-songsee)、[`youtube-content`](/user-guide/skills/bundled/media/media-youtube-content)、[`claude-design`](/user-guide/skills/bundled/creative/creative-claude-design)、[`excalidraw`](/user-guide/skills/bundled/creative/creative-excalidraw)、[`architecture-diagram`](/user-guide/skills/bundled/creative/creative-architecture-diagram)、[`concept-diagrams`](/user-guide/skills/optional/creative/creative-concept-diagrams)、[`baoyu-comic`](/user-guide/skills/bundled/creative/creative-baoyu-comic)、[`baoyu-infographic`](/user-guide/skills/bundled/creative/creative-baoyu-infographic)、[`humanizer`](/user-guide/skills/bundled/creative/creative-humanizer)、[`gif-search`](/user-guide/skills/bundled/media/media-gif-search)、[`meme-generation`](/user-guide/skills/optional/creative/creative-meme-generation) |
## 参考:完整 SKILL.md
@ -41,7 +41,7 @@ description: "规划、搭建并监控由 Hermes Kanban 支撑的多智能体视
4. **交接**——移交给 director profile由其通过 kanban 进行分解
5. **监控**——跟踪执行过程,在任务卡住或失败时协助介入
实际渲染在 kanban 运行后在其内部完成,使用适合各场景的现有技能和工具——`ascii-video`、`manim-video`、`p5js`、`comfyui`、`touchdesigner-mcp`、`blender-mcp`、`songwriting-and-ai-music`、`heartmula`、外部 API或使用 PIL + ffmpeg 的纯 Python。
实际渲染在 kanban 运行后在其内部完成,使用适合各场景的现有技能和工具——`ascii-video`、`manim-video`、`p5js`、`comfyui`、`touchdesigner-mcp`、`songwriting-and-ai-music`、`heartmula`、外部 API或使用 PIL + ffmpeg 的纯 Python。
## 不适用本技能的情况

View File

@ -376,7 +376,6 @@ const sidebars: SidebarsConfig = {
'user-guide/skills/optional/creative/creative-audiocraft-audio-generation',
'user-guide/skills/optional/creative/creative-baoyu-article-illustrator',
'user-guide/skills/optional/creative/creative-baoyu-comic',
'user-guide/skills/optional/creative/creative-blender-mcp',
'user-guide/skills/optional/creative/creative-concept-diagrams',
'user-guide/skills/optional/creative/creative-creative-ideation',
'user-guide/skills/optional/creative/creative-heartmula',