[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EAGGC80T2WXF44VV3YCZWN":3,"skill-same:01M7EAGGC80T2WXF44VV3YCZWN":89},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":45,"commitSha":46,"skillDirectory":47,"forksCount":48,"repositoryHtmlUrl":49,"repositoryHomepageUrl":6,"ownerHtmlUrl":50,"ownerBlogUrl":6,"ownerDescription":51,"updatedAt":52,"install":53,"resources":66},"01M7EAGGC80T2WXF44VV3YCZWN","unreal-mcp",null,"Automate Unreal Engine editor scenes, actors, and renders.","sanjeevafk\u002Fagent-skills","sanjeevafk","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F230678117?v=4","01M7C540EFY46ETG1NNWKS5X3Q","Sanjeev","01M7C540F366J5C1D587BMKH9Q",0,"MIT","https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fcollections\u002Fhermes-agent\u002Fskills\u002Funreal-mcp\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: unreal-mcp\ndescription: Automate Unreal Engine editor scenes, actors, and renders.\nversion: 1.0.0\nrequires: Unreal Editor 5.8+ with the Unreal MCP plugin enabled and its server running\nauthor: Hermes Agent\nlicense: MIT\ntags: [unreal, unreal-engine, ue5, 3d, mcp, scenes, cinematics, lighting, gamedev]\nplatforms: [linux, macos, windows]\nmetadata:\n  hermes:\n    tags: [unreal, unreal-engine, ue5, 3d, mcp, scenes, cinematics, lighting, gamedev]\n    related_skills: []\n---\n\n# Unreal Engine MCP Skill\n\nCompanion skill for the `unreal-engine` entry in the Hermes MCP catalog. The\nMCP server (Epic's official, experimental \"Unreal MCP\" plugin, internal id\n`ModelContextProtocol`) runs INSIDE the Unreal Editor process and exposes\neditor functionality as typed tools. This skill teaches how to drive it well:\ndiscovering the live tool surface, sequencing calls safely, translating\nplain-English asks into scenes that actually look good, and verifying work\nvisually. The user should never need to touch the editor beyond launching it.\n\n## When to Use\n\nUse when the user wants anything done in Unreal Engine: build or dress a\nlevel, spawn\u002Fmove\u002Fdelete actors, set up lighting and atmosphere, create or\ntune material instances, frame a camera shot, capture screenshots or renders,\nimport assets, inspect the scene or UI, run automation tests, or script the\neditor. Works for single actions (\"make the sun golden hour\") and for\ncomplete multi-step projects (\"build me a moody forest clearing with a\ncampfire and render a shot of it\").\n\nDon't use for: DCC-style mesh modeling\u002Fsculpting (model in Blender and\nimport the result), or for editing Unreal C++ project source (that's normal\ncode work — use the terminal; this skill is about the live editor).\n\n## Prerequisites\n\nTwo halves, in this order: the editor side must be up before Hermes connects.\n\n### One-time, editor side\n\n1. Unreal Editor **5.8+** with a project open. (macOS: full Xcode must be\n   installed and its license accepted — the editor exits on first launch\n   without it; see pitfalls.)\n2. **Edit > Plugins** — enable **Unreal MCP** (its Toolset Registry\n   dependency auto-enables). Restart the editor when prompted.\n3. The typed toolsets ship separately from the server: also enable the\n   **AllToolsets** plugin in the same Plugins browser. Unreal MCP ships NO\n   tools itself — AllToolsets provides the shipped toolsets (SceneTools,\n   ActorTools, MaterialInstanceTools, ObjectTools, …); skip it and the\n   server connects but the agent has nothing to call.\n4. **Edit > Editor Preferences > General > Model Context Protocol** — enable\n   **Auto Start Server**. Default bind is `http:\u002F\u002F127.0.0.1:8000\u002Fmcp`\n   (port\u002Fpath configurable in the same panel; server name is `unreal-mcp`).\n   To start manually instead, run `ModelContextProtocol.StartServer` in the\n   editor console (backtick key).\n\n### One-time, Hermes side\n\n    hermes mcp install unreal-engine\n\nThis writes the `mcp_servers.unreal-engine` HTTP entry pointing at\n`http:\u002F\u002F127.0.0.1:8000\u002Fmcp` and probes the live server for its tools. Run it\nwhile the editor + server are up so the probe sees the real surface. If the\nuser changed port\u002Fpath in Editor Preferences, edit the `url` in\n`~\u002F.hermes\u002Fconfig.yaml` under `mcp_servers.unreal-engine` to match.\n\nDo NOT use `ModelContextProtocol.GenerateClientConfig` for Hermes — that\nwrites `.mcp.json`-style files for Claude Code\u002FCursor\u002Fetc. Hermes connects\nfrom `config.yaml` via the catalog entry.\n\n### Every session\n\n1. Launch Unreal Editor, wait for the project to finish loading; confirm the\n   server started (Output Log shows the bind address, or run\n   `ModelContextProtocol.StartServer` manually).\n2. Start the Hermes session. Tools register as `mcp_unreal_engine_*`. If\n   they're missing: editor wasn't up first — start it, then open a new\n   Hermes session.\n3. Sanity check: call `mcp_unreal_engine_list_toolsets` and confirm toolsets\n   come back.\n\n## The Tool Surface: Discovery, Not a Fixed List\n\nBy default the plugin runs in **tool-search mode**: `tools\u002Flist` returns only\nthree meta-tools, and every real tool is reached through them. Through Hermes\nthey appear as:\n\n| Hermes tool | Purpose |\n|---|---|\n| `mcp_unreal_engine_list_toolsets` | Names + descriptions of every registered toolset |\n| `mcp_unreal_engine_describe_toolset` | Full JSON schemas for one named toolset's tools |\n| `mcp_unreal_engine_call_tool` | Invoke a named tool with arguments, get the result |\n\nThe discovery walk, always in this order:\n\n1. `list_toolsets` → see what capability groups this project actually has\n   (the surface is project-dependent: enabled plugins, Game Feature Plugins,\n   and any custom toolsets all contribute). Names come back FULLY QUALIFIED\n   (`editor_toolset.toolsets.scene.SceneTools`,\n   `EditorToolset.EditorAppToolset`) — use them verbatim as `toolset_name`.\n2. `describe_toolset` on the group you need → read the real parameter\n   schemas. Never guess parameter names — schemas are the contract.\n3. `call_tool` with the qualified toolset name, the SHORT tool name\n   (`find_actors`, not the dotted form), and arguments matching the schema.\n\nCache what you learn for the session; re-list only after the editor side\nchanges (new plugin enabled, toolset authored, `RefreshTools` run).\n\nThe alternative eager mode (`Enable Tool Search` off in Editor Preferences)\nadvertises every tool as its own `mcp_unreal_engine_{tool}` entry. Discovery\nthen happens at `hermes mcp install`\u002F`configure` time instead. Tool-search\nmode is the default and what this skill assumes; it also keeps schema tokens\nout of every API call, so prefer it.\n\nSee `references\u002Ftool-surface.md` for the shipped toolset catalog, authoring\ncustom toolsets, and the full plugin configuration\u002Fconsole-command reference.\n\n## Operating Loop\n\nEvery Unreal task follows the same loop:\n\n1. **Inspect first.** List toolsets, then query the scene\u002Flevel state before\n   touching anything. Never assume an empty or default level. In an\n   unfamiliar project, also check for project-registered Agent Skills\n   (`call_tool` → `AgentSkillToolset.ListSkills`): a matching project skill's\n   instructions override this skill's generic defaults.\n2. **Act in small, single-purpose calls.** One logical step per `call_tool`.\n   The server executes tools **serially on the game thread** — a big\n   monolithic operation freezes the editor UI until it finishes and risks\n   client timeouts. Exception: for loops over 5+ homogeneous operations,\n   ONE `ProgrammaticToolset.execute_tool_script` call batches them\n   server-side without breaking the serial rule\n   (`references\u002Fadvanced-workflows.md`).\n3. **NEVER issue overlapping calls.** Do not batch multiple\n   `mcp_unreal_engine_*` calls in one turn — Hermes runs batched calls\n   concurrently, and parallel calls against the game thread deadlock or\n   fail. Strictly one call, await result, next call. This overrides the\n   general parallel-tool-calls guidance.\n4. **Read every result.** Many tools (Blueprint compiles, material edits,\n   widget creation) report success\u002Ffailure in the response body with no\n   protocol-level exception. Anything that isn't an explicit success is a\n   stop-and-diagnose, not a shrug. After property writes, read the value\n   back — several write paths silently no-op (see pitfalls).\n5. **Verify visually and structurally.** After each milestone, confirm state\n   by querying the actors\u002Fproperties you changed, and capture a viewport\n   screenshot when composition matters (see `references\u002Ftool-surface.md` for\n   the capture options; `vision_analyze` the image — you are the art\n   director, judge it).\n6. **Save often.** Editor edits are in-memory until packages\u002Flevels are\n   saved; an editor crash loses everything since the last save, and MCP\n   edits are not reliably undoable. Save before AND after any bulk change,\n   and after every milestone.\n7. **Report concretely.** Actor labels, asset paths (`\u002FGame\u002F...`), file\n   locations of captures\u002Frenders.\n\nRules of the world while you work:\n\n- Units are **centimeters**; axes are **Z-up**, X-forward; rotations are\n  degrees (Rotator: Roll around X, Pitch around Y, Yaw around Z). Human eye\n  height ≈ 165 cm; a door ≈ 210×90 cm. Full tables in\n  `references\u002Fscene-craft.md`.\n- Content paths use long package names: `\u002FGame\u002FFolder\u002FAsset.Asset` for\n  project content, `\u002FEngine\u002FBasicShapes\u002FCube.Cube` for engine primitives.\n- Actor **labels** (what you see in the Outliner, settable, non-unique) are\n  not actor **names** (internal, unique). Prefer resolving actors by\n  label\u002Fclass queries, then hold on to whatever handle the tool returns.\n- Prefer physically-plausible lighting values (lux\u002Fcandela\u002FKelvin) over\n  arbitrary brightness numbers — but FIRST read the existing sun's\n  intensity to learn the scene's calibration convention; template worlds\n  are often calibrated around `intensity: 10`, and physical values blow\n  them out (`references\u002Fscene-craft.md` has the numbers,\n  `references\u002Fpitfalls.md` #12b has the calibration rule).\n\n## From Plain English to a Scene\n\nThe user gives intent, not specs. Translate before you build:\n\n1. **Extract the brief.** Subject, mood, time of day, interior\u002Fexterior,\n   style, deliverable (screenshot? render? playable level?). Ask at most one\n   round of clarifying questions, then commit — you are the technical\n   director; don't bounce Unreal jargon back at the user.\n2. **Plan the build order.** The order that works: level\u002Fenvironment shell →\n   blocking (major geometry\u002Fmeshes in place) → lighting + atmosphere →\n   materials → set dressing\u002Fdetail → camera → capture\u002Frender. Post the plan\n   as a todo list for multi-step builds.\n3. **Build with the loop above**, one milestone at a time, screenshot at\n   each milestone.\n4. **Art-direct yourself.** Compare each screenshot against the brief:\n   readable silhouette? believable light direction\u002Fintensity? horizon not\n   dead-center? scale correct against a human-height reference? Fix before\n   moving on.\n5. **Deliver.** Screenshots\u002Frenders as files (`MEDIA:` path), plus a short\n   summary of what exists in the level and where it was saved.\n\n`references\u002Frecipes.md` has complete worked builds (exterior daylight scene,\nmoody interior, golden-hour cinematic + render, asset import & placement)\nwith the exact call sequences and values.\n\n## Reference Files\n\nLoad on demand; keep SKILL.md-level rules in mind throughout.\n\n| Reference | Contents |\n|---|---|\n| `references\u002Ftool-surface.md` | Shipped toolsets catalog, discovery protocol detail, plugin console commands\u002FCVars\u002Fflags, screenshot & capture paths, MCP Inspector debugging, extending with custom Python\u002FC++ toolsets |\n| `references\u002Fadvanced-workflows.md` | Sophisticated workflows, live-verified: ProgrammaticToolset batching, Blueprint DSL authoring loop (create→DSL→compile→spawn), PIE test sessions, Sequencer orientation (140 tools), LogsToolset self-debugging, automation testing, semantic asset search, config settings, per-situation decision table |\n| `references\u002Fscene-craft.md` | Numeric cheat sheet: physical light intensities, color temperatures, exposure\u002FEV100, fog densities, mood recipes (noon\u002Fgolden hour\u002Fovercast\u002Fnight\u002Finterior), scale tables, content path conventions |\n| `references\u002Frecipes.md` | End-to-end worked builds with exact call sequences |\n| `references\u002Fpitfalls.md` | Setup, runtime, and workflow pitfalls with fixes — read before your first session and whenever something misbehaves |\n\n## Pitfalls (top of mind — full list in references\u002Fpitfalls.md)\n\n- **Start order matters.** Editor + server up first, then the Hermes\n  session. Missing `mcp_unreal_engine_*` tools = wrong order.\n- **One call at a time.** Serial game thread; no batching, no overlap.\n- **The editor UI freezes during each call.** That's by design (game-thread\n  execution). Warn the user during long operations; keep calls small.\n- **Modal dialogs block everything.** A tool call that opens (or collides\n  with) a modal editor dialog stalls until a human dismisses it. If a call\n  hangs indefinitely, tell the user to check the editor for a dialog.\n- **Timeouts on long operations.** Hermes' per-call default is 120 s; asset\n  imports, big level saves, and renders can exceed it. Raise\n  `mcp_servers.unreal-engine.timeout` in `~\u002F.hermes\u002Fconfig.yaml` for\n  render\u002Fimport-heavy sessions.\n- **Stale tool schemas.** After authoring\u002Fhot-reloading toolsets or enabling\n  a plugin, run `ModelContextProtocol.RefreshTools` in the editor console\n  and re-`list_toolsets`. New C++ `UFUNCTION`s need a full editor restart —\n  Live Coding won't surface them.\n- **Experimental plugin.** APIs and tool shapes can change between engine\n  versions; trust `describe_toolset` over memory, including this skill's\n  examples. When docs and the live schema disagree, the live schema wins.\n- **Don't expose the server beyond localhost.** Loopback-only, no auth, by\n  design. Never suggest binding it wider.\n- **Licensing note.** The server logs on start: data transmitted via the\n  plugin to a connected LLM service is Licensed Technology under the UE\n  EULA (§6(e)) — the user is responsible for ensuring their LLM provider\n  doesn't train on it. Surface this if the user asks about data handling.\n\n## Verification Checklist\n\n- [ ] `list_toolsets` returns toolsets at session start (connection healthy)\n- [ ] Scene state queried before first edit (never assumed empty)\n- [ ] After each milestone: changed actors\u002Fproperties re-queried and a\n      screenshot reviewed against the brief\n- [ ] Level\u002Fdirty packages saved after each milestone and at the end\n- [ ] Deliverables exist on disk (screenshot\u002Frender paths confirmed) and are\n      reported to the user with absolute paths\n- [ ] Editor left in a clean state: no pending modal, no unsaved surprise,\n      user told exactly what was created\u002Fchanged and where\n","# Unreal Engine MCP Skill\n\nCompanion skill for the `unreal-engine` entry in the Hermes MCP catalog. The\nMCP server (Epic's official, experimental \"Unreal MCP\" plugin, internal id\n`ModelContextProtocol`) runs INSIDE the Unreal Editor process and exposes\neditor functionality as typed tools. This skill teaches how to drive it well:\ndiscovering the live tool surface, sequencing calls safely, translating\nplain-English asks into scenes that actually look good, and verifying work\nvisually. The user should never need to touch the editor beyond launching it.\n\n## When to Use\n\nUse when the user wants anything done in Unreal Engine: build or dress a\nlevel, spawn\u002Fmove\u002Fdelete actors, set up lighting and atmosphere, create or\ntune material instances, frame a camera shot, capture screenshots or renders,\nimport assets, inspect the scene or UI, run automation tests, or script the\neditor. Works for single actions (\"make the sun golden hour\") and for\ncomplete multi-step projects (\"build me a moody forest clearing with a\ncampfire and render a shot of it\").\n\nDon't use for: DCC-style mesh modeling\u002Fsculpting (model in Blender and\nimport the result), or for editing Unreal C++ project source (that's normal\ncode work — use the terminal; this skill is about the live editor).\n\n## Prerequisites\n\nTwo halves, in this order: the editor side must be up before Hermes connects.\n\n### One-time, editor side\n\n1. Unreal Editor **5.8+** with a project open. (macOS: full Xcode must be\n   installed and its license accepted — the editor exits on first launch\n   without it; see pitfalls.)\n2. **Edit > Plugins** — enable **Unreal MCP** (its Toolset Registry\n   dependency auto-enables). Restart the editor when prompted.\n3. The typed toolsets ship separately from the server: also enable the\n   **AllToolsets** plugin in the same Plugins browser. Unreal MCP ships NO\n   tools itself — AllToolsets provides the shipped toolsets (SceneTools,\n   ActorTools, MaterialInstanceTools, ObjectTools, …); skip it and the\n   server connects but the agent has nothing to call.\n4. **Edit > Editor Preferences > General > Model Context Protocol** — enable\n   **Auto Start Server**. Default bind is `http:\u002F\u002F127.0.0.1:8000\u002Fmcp`\n   (port\u002Fpath configurable in the same panel; server name is `unreal-mcp`).\n   To start manually instead, run `ModelContextProtocol.StartServer` in the\n   editor console (backtick key).\n\n### One-time, Hermes side\n\n    hermes mcp install unreal-engine\n\nThis writes the `mcp_servers.unreal-engine` HTTP entry pointing at\n`http:\u002F\u002F127.0.0.1:8000\u002Fmcp` and probes the live server for its tools. Run it\nwhile the editor + server are up so the probe sees the real surface. If the\nuser changed port\u002Fpath in Editor Preferences, edit the `url` in\n`~\u002F.hermes\u002Fconfig.yaml` under `mcp_servers.unreal-engine` to match.\n\nDo NOT use `ModelContextProtocol.GenerateClientConfig` for Hermes — that\nwrites `.mcp.json`-style files for Claude Code\u002FCursor\u002Fetc. Hermes connects\nfrom `config.yaml` via the catalog entry.\n\n### Every session\n\n1. Launch Unreal Editor, wait for the project to finish loading; confirm the\n   server started (Output Log shows the bind address, or run\n   `ModelContextProtocol.StartServer` manually).\n2. Start the Hermes session. Tools register as `mcp_unreal_engine_*`. If\n   they're missing: editor wasn't up first — start it, then open a new\n   Hermes session.\n3. Sanity check: call `mcp_unreal_engine_list_toolsets` and confirm toolsets\n   come back.\n\n## The Tool Surface: Discovery, Not a Fixed List\n\nBy default the plugin runs in **tool-search mode**: `tools\u002Flist` returns only\nthree meta-tools, and every real tool is reached through them. Through Hermes\nthey appear as:\n\n| Hermes tool | Purpose |\n|---|---|\n| `mcp_unreal_engine_list_toolsets` | Names + descriptions of every registered toolset |\n| `mcp_unreal_engine_describe_toolset` | Full JSON schemas for one named toolset's tools |\n| `mcp_unreal_engine_call_tool` | Invoke a named tool with arguments, get the result |\n\nThe discovery walk, always in this order:\n\n1. `list_toolsets` → see what capability groups this project actually has\n   (the surface is project-dependent: enabled plugins, Game Feature Plugins,\n   and any custom toolsets all contribute). Names come back FULLY QUALIFIED\n   (`editor_toolset.toolsets.scene.SceneTools`,\n   `EditorToolset.EditorAppToolset`) — use them verbatim as `toolset_name`.\n2. `describe_toolset` on the group you need → read the real parameter\n   schemas. Never guess parameter names — schemas are the contract.\n3. `call_tool` with the qualified toolset name, the SHORT tool name\n   (`find_actors`, not the dotted form), and arguments matching the schema.\n\nCache what you learn for the session; re-list only after the editor side\nchanges (new plugin enabled, toolset authored, `RefreshTools` run).\n\nThe alternative eager mode (`Enable Tool Search` off in Editor Preferences)\nadvertises every tool as its own `mcp_unreal_engine_{tool}` entry. Discovery\nthen happens at `hermes mcp install`\u002F`configure` time instead. Tool-search\nmode is the default and what this skill assumes; it also keeps schema tokens\nout of every API call, so prefer it.\n\nSee `references\u002Ftool-surface.md` for the shipped toolset catalog, authoring\ncustom toolsets, and the full plugin configuration\u002Fconsole-command reference.\n\n## Operating Loop\n\nEvery Unreal task follows the same loop:\n\n1. **Inspect first.** List toolsets, then query the scene\u002Flevel state before\n   touching anything. Never assume an empty or default level. In an\n   unfamiliar project, also check for project-registered Agent Skills\n   (`call_tool` → `AgentSkillToolset.ListSkills`): a matching project skill's\n   instructions override this skill's generic defaults.\n2. **Act in small, single-purpose calls.** One logical step per `call_tool`.\n   The server executes tools **serially on the game thread** — a big\n   monolithic operation freezes the editor UI until it finishes and risks\n   client timeouts. Exception: for loops over 5+ homogeneous operations,\n   ONE `ProgrammaticToolset.execute_tool_script` call batches them\n   server-side without breaking the serial rule\n   (`references\u002Fadvanced-workflows.md`).\n3. **NEVER issue overlapping calls.** Do not batch multiple\n   `mcp_unreal_engine_*` calls in one turn — Hermes runs batched calls\n   concurrently, and parallel calls against the game thread deadlock or\n   fail. Strictly one call, await result, next call. This overrides the\n   general parallel-tool-calls guidance.\n4. **Read every result.** Many tools (Blueprint compiles, material edits,\n   widget creation) report success\u002Ffailure in the response body with no\n   protocol-level exception. Anything that isn't an explicit success is a\n   stop-and-diagnose, not a shrug. After property writes, read the value\n   back — several write paths silently no-op (see pitfalls).\n5. **Verify visually and structurally.** After each milestone, confirm state\n   by querying the actors\u002Fproperties you changed, and capture a viewport\n   screenshot when composition matters (see `references\u002Ftool-surface.md` for\n   the capture options; `vision_analyze` the image — you are the art\n   director, judge it).\n6. **Save often.** Editor edits are in-memory until packages\u002Flevels are\n   saved; an editor crash loses everything since the last save, and MCP\n   edits are not reliably undoable. Save before AND after any bulk change,\n   and after every milestone.\n7. **Report concretely.** Actor labels, asset paths (`\u002FGame\u002F...`), file\n   locations of captures\u002Frenders.\n\nRules of the world while you work:\n\n- Units are **centimeters**; axes are **Z-up**, X-forward; rotations are\n  degrees (Rotator: Roll around X, Pitch around Y, Yaw around Z). Human eye\n  height ≈ 165 cm; a door ≈ 210×90 cm. Full tables in\n  `references\u002Fscene-craft.md`.\n- Content paths use long package names: `\u002FGame\u002FFolder\u002FAsset.Asset` for\n  project content, `\u002FEngine\u002FBasicShapes\u002FCube.Cube` for engine primitives.\n- Actor **labels** (what you see in the Outliner, settable, non-unique) are\n  not actor **names** (internal, unique). Prefer resolving actors by\n  label\u002Fclass queries, then hold on to whatever handle the tool returns.\n- Prefer physically-plausible lighting values (lux\u002Fcandela\u002FKelvin) over\n  arbitrary brightness numbers — but FIRST read the existing sun's\n  intensity to learn the scene's calibration convention; template worlds\n  are often calibrated around `intensity: 10`, and physical values blow\n  them out (`references\u002Fscene-craft.md` has the numbers,\n  `references\u002Fpitfalls.md` #12b has the calibration rule).\n\n## From Plain English to a Scene\n\nThe user gives intent, not specs. Translate before you build:\n\n1. **Extract the brief.** Subject, mood, time of day, interior\u002Fexterior,\n   style, deliverable (screenshot? render? playable level?). Ask at most one\n   round of clarifying questions, then commit — you are the technical\n   director; don't bounce Unreal jargon back at the user.\n2. **Plan the build order.** The order that works: level\u002Fenvironment shell →\n   blocking (major geometry\u002Fmeshes in place) → lighting + atmosphere →\n   materials → set dressing\u002Fdetail → camera → capture\u002Frender. Post the plan\n   as a todo list for multi-step builds.\n3. **Build with the loop above**, one milestone at a time, screenshot at\n   each milestone.\n4. **Art-direct yourself.** Compare each screenshot against the brief:\n   readable silhouette? believable light direction\u002Fintensity? horizon not\n   dead-center? scale correct against a human-height reference? Fix before\n   moving on.\n5. **Deliver.** Screenshots\u002Frenders as files (`MEDIA:` path), plus a short\n   summary of what exists in the level and where it was saved.\n\n`references\u002Frecipes.md` has complete worked builds (exterior daylight scene,\nmoody interior, golden-hour cinematic + render, asset import & placement)\nwith the exact call sequences and values.\n\n## Reference Files\n\nLoad on demand; keep SKILL.md-level rules in mind throughout.\n\n| Reference | Contents |\n|---|---|\n| `references\u002Ftool-surface.md` | Shipped toolsets catalog, discovery protocol detail, plugin console commands\u002FCVars\u002Fflags, screenshot & capture paths, MCP Inspector debugging, extending with custom Python\u002FC++ toolsets |\n| `references\u002Fadvanced-workflows.md` | Sophisticated workflows, live-verified: ProgrammaticToolset batching, Blueprint DSL authoring loop (create→DSL→compile→spawn), PIE test sessions, Sequencer orientation (140 tools), LogsToolset self-debugging, automation testing, semantic asset search, config settings, per-situation decision table |\n| `references\u002Fscene-craft.md` | Numeric cheat sheet: physical light intensities, color temperatures, exposure\u002FEV100, fog densities, mood recipes (noon\u002Fgolden hour\u002Fovercast\u002Fnight\u002Finterior), scale tables, content path conventions |\n| `references\u002Frecipes.md` | End-to-end worked builds with exact call sequences |\n| `references\u002Fpitfalls.md` | Setup, runtime, and workflow pitfalls with fixes — read before your first session and whenever something misbehaves |\n\n## Pitfalls (top of mind — full list in references\u002Fpitfalls.md)\n\n- **Start order matters.** Editor + server up first, then the Hermes\n  session. Missing `mcp_unreal_engine_*` tools = wrong order.\n- **One call at a time.** Serial game thread; no batching, no overlap.\n- **The editor UI freezes during each call.** That's by design (game-thread\n  execution). Warn the user during long operations; keep calls small.\n- **Modal dialogs block everything.** A tool call that opens (or collides\n  with) a modal editor dialog stalls until a human dismisses it. If a call\n  hangs indefinitely, tell the user to check the editor for a dialog.\n- **Timeouts on long operations.** Hermes' per-call default is 120 s; asset\n  imports, big level saves, and renders can exceed it. Raise\n  `mcp_servers.unreal-engine.timeout` in `~\u002F.hermes\u002Fconfig.yaml` for\n  render\u002Fimport-heavy sessions.\n- **Stale tool schemas.** After authoring\u002Fhot-reloading toolsets or enabling\n  a plugin, run `ModelContextProtocol.RefreshTools` in the editor console\n  and re-`list_toolsets`. New C++ `UFUNCTION`s need a full editor restart —\n  Live Coding won't surface them.\n- **Experimental plugin.** APIs and tool shapes can change between engine\n  versions; trust `describe_toolset` over memory, including this skill's\n  examples. When docs and the live schema disagree, the live schema wins.\n- **Don't expose the server beyond localhost.** Loopback-only, no auth, by\n  design. Never suggest binding it wider.\n- **Licensing note.** The server logs on start: data transmitted via the\n  plugin to a connected LLM service is Licensed Technology under the UE\n  EULA (§6(e)) — the user is responsible for ensuring their LLM provider\n  doesn't train on it. Surface this if the user asks about data handling.\n\n## Verification Checklist\n\n- [ ] `list_toolsets` returns toolsets at session start (connection healthy)\n- [ ] Scene state queried before first edit (never assumed empty)\n- [ ] After each milestone: changed actors\u002Fproperties re-queried and a\n      screenshot reviewed against the brief\n- [ ] Level\u002Fdirty packages saved after each milestone and at the end\n- [ ] Deliverables exist on disk (screenshot\u002Frender paths confirmed) and are\n      reported to the user with absolute paths\n- [ ] Editor left in a clean state: no pending modal, no unsaved surprise,\n      user told exactly what was created\u002Fchanged and where\n",{"name":5,"tags":24,"author":34,"license":15,"version":35,"metadata":36,"requires":40,"platforms":41,"description":7},[25,26,27,28,29,30,31,32,33],"unreal","unreal-engine","ue5","3d","mcp","scenes","cinematics","lighting","gamedev","Hermes Agent","1.0.0",{"hermes":37},{"tags":38,"related_skills":39},[25,26,27,28,29,30,31,32,33],[],"Unreal Editor 5.8+ with the Unreal MCP plugin enabled and its server running",[42,43,44],"linux","macos","windows","main","4e084e84eb42fc53e144c89c0f7f48b39c8a4b0f","collections\u002Fhermes-agent\u002Fskills\u002Funreal-mcp",1,"https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills","https:\u002F\u002Fgithub.com\u002Fsanjeevafk","First make it work. Then make it right. Then make it fast.","2026-11-14 13:36:16",{"source":8,"skillName":5,"commands":54,"prompt":64,"zipAvailable":65},[55,58,61],{"tool":56,"command":57},"npx","npx skills add sanjeevafk\u002Fagent-skills --skill unreal-mcp",{"tool":59,"command":60},"bunx","bunx skills add sanjeevafk\u002Fagent-skills --skill unreal-mcp",{"tool":62,"command":63},"pnpm","pnpm dlx skills add sanjeevafk\u002Fagent-skills --skill unreal-mcp","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fcollections\u002Fhermes-agent\u002Fskills\u002Funreal-mcp\u002FSKILL.md",true,[67,73,77,81,85],{"path":68,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":71,"sourceHtmlUrl":72},"references\u002Fadvanced-workflows.md","REFERENCE","参考文档",13003,"https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fcollections\u002Fhermes-agent\u002Fskills\u002Funreal-mcp\u002Freferences\u002Fadvanced-workflows.md",{"path":74,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":75,"sourceHtmlUrl":76},"references\u002Fpitfalls.md",17635,"https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fcollections\u002Fhermes-agent\u002Fskills\u002Funreal-mcp\u002Freferences\u002Fpitfalls.md",{"path":78,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":79,"sourceHtmlUrl":80},"references\u002Frecipes.md",10609,"https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fcollections\u002Fhermes-agent\u002Fskills\u002Funreal-mcp\u002Freferences\u002Frecipes.md",{"path":82,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":83,"sourceHtmlUrl":84},"references\u002Fscene-craft.md",12793,"https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fcollections\u002Fhermes-agent\u002Fskills\u002Funreal-mcp\u002Freferences\u002Fscene-craft.md",{"path":86,"resourceType":69,"resourceTypeDesc":70,"sizeBytes":87,"sourceHtmlUrl":88},"references\u002Ftool-surface.md",19446,"https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fcollections\u002Fhermes-agent\u002Fskills\u002Funreal-mcp\u002Freferences\u002Ftool-surface.md",[90,97,105,114,122,129],{"skillNo":91,"name":92,"nameZh":6,"description":93,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":96,"publishedAt":19},"01M7EFHX67M3KKXT1BXVPA5ADT","zarr-python","Chunked N-D arrays for cloud storage (Zarr-Python 3). Compressed arrays, parallel I\u002FO, S3\u002FGCS via fsspec, NumPy\u002FDask\u002FXarray compatible, for large-scale scientific computing pipelines.","https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzarr-python\u002FSKILL.md",[],[],{"skillNo":98,"name":99,"nameZh":6,"description":100,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":101,"sourceHtmlUrl":102,"categories":103,"keywords":104,"publishedAt":19},"01M7EFFQGM51TZBGRENXY50ZYM","whiteboard-diagrams","Create accurate, simple, hand-drawn whiteboard diagrams from a user's process, system, architecture, learning workflow, comparison, or causal explanation. Use when the user asks for a scribbly whiteboard visual, educational workflow graphic, hand-drawn explainer, or diagram like the bundled reference; ask one focused clarification question when the subject or relationships are missing, then generate and visually verify the final image.","https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwhiteboard-diagrams\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwhiteboard-diagrams\u002FSKILL.md",[],[],{"skillNo":106,"name":107,"nameZh":6,"description":108,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":109,"licenseFileUrl":110,"sourceHtmlUrl":111,"categories":112,"keywords":113,"publishedAt":19},"01M7EFF1CSQNTDMHFK4ZGY6NYK","webapp-testing","Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.","Apache-2.0","https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwebapp-testing\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwebapp-testing\u002FSKILL.md",[],[],{"skillNo":115,"name":116,"nameZh":6,"description":117,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":109,"licenseFileUrl":118,"sourceHtmlUrl":119,"categories":120,"keywords":121,"publishedAt":19},"01M7EFE8106X6C28FT8PA74427","web-artifacts-builder","Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn\u002Fui). Use for complex artifacts requiring state management, routing, or shadcn\u002Fui components - not for simple single-file HTML\u002FJSX artifacts.","https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fweb-artifacts-builder\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fweb-artifacts-builder\u002FSKILL.md",[],[],{"skillNo":123,"name":124,"nameZh":6,"description":125,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":126,"categories":127,"keywords":128,"publishedAt":19},"01M7EFCSNZM9G7ZHS6F127Q71X","verification-loop","A comprehensive verification system for Claude Code sessions.","https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fverification-loop\u002FSKILL.md",[],[],{"skillNo":130,"name":131,"nameZh":6,"description":132,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":133,"sourceHtmlUrl":134,"categories":135,"keywords":136,"publishedAt":19},"01M7EFCE7EZMKPRYDDNKBQ9EEH","vercel-deploy","Deploy applications and websites to Vercel. Use when the user requests deployment actions like \"deploy my app\", \"deploy and give me the link\", \"push this live\", or \"create a preview deployment\".","https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-deploy\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fsanjeevafk\u002Fagent-skills\u002Fblob\u002Fmain\u002Fskills\u002Fvercel-deploy\u002FSKILL.md",[],[]]