Como usar
- Copie o comando e verifique o pacote antes de instalar no seu ambiente de agente.
- Consulte no original as ferramentas, credenciais e ambientes necessários.
- Descreva os dados de entrada e a entrega esperada; confira a saída com as fontes.
npx skills add https://github.com/thanhnguyxnorg/blendops --skill product-hero-scene-plannerAntes de começar
- A instalação não inclui assinatura de modelo, franquia API ou acesso aos seus arquivos.
- Estrelas e instalações indicam atividade, não um teste independente de qualidade ou segurança.
Renderizado do repositório de origem, preservando títulos, exemplos, código, tabelas, links e imagens.
product-hero-scene-planner
Purpose
Turn user intent or an upstream brief into a complete, plain-language product-hero scene plan (subject, composition, camera, lighting, material, environment, plus render, export, and web-handoff intent) before any runtime phase.
Canonical role: high-level scene-planning for single product-hero workflows in BlendOps. Keep role boundaries strict:
intent-to-3d-brief-writerwrites the upstream brief.- THIS skill (
product-hero-scene-planner) produces the complete product-hero scene plan and assumptions register. camera-compositiondeep-dives framing and lens setup.lighting-material-plandeep-dives lookdev and shader/light rigs.acceptance-criteria-builderwrites measurable acceptance rows.done-criteriapins the done condition.scene-readinessreviews scene readiness later.
Quick start
- confirm this skill fits your goal
- provide required inputs first
- keep runtime claims evidence-bound
- follow suggested next-skill handoff
When to use
- single product-hero scene planning from user intent or an
intent-to-3d-brief-writerbrief - pre-runtime structural planning across subject, camera, composition, lighting, material, environment, and delivery intent
- translating non-Blender product presentation goals into an assumption-tracked scene plan
When not to use
<!-- blendops:generated:start exclusions -->
- to execute Blender commands, write Python, or claim generated artifacts → use
../intent-to-3d-brief-writer/SKILL.md(Delegates out-of-scope requests to intent-to-3d-brief-writer) - to skip unresolved critical ambiguity or hide missing inputs → use
../camera-composition/SKILL.md(Delegates out-of-scope requests to camera-composition) - when raw user intent still needs initial brief normalization → use
../intent-to-3d-brief-writer/SKILL.md(use intent-to-3d-brief-writer first) - when planning a multi-item catalog or grid layout → use
../product-grid-scene-planner/SKILL.md(use product-grid-scene-planner) - when deep-diving camera/framing mechanics → use
../camera-composition/SKILL.md(use lighting-material-plan) - when writing measurable pass/fail acceptance tables → use
../acceptance-criteria-builder/SKILL.md(use scene-readiness)
<!-- blendops:generated:end exclusions -->
Trigger phrases
<!-- blendops:generated:start triggers -->
- "plan a product hero scene"
- "turn this into a Blender plan"
- "prepare cyberpunk shoe scene plan"
<!-- blendops:generated:end triggers -->
Prerequisites / readiness
- output destination clarified (web, app, game, or still/video render)
- upstream brief from
intent-to-3d-brief-writerprovided or user intent clear enough to extract constraints - runtime readiness optionally checked via
runtime-readiness(does not alter this skill's truth states: Runtime staysNot Run, Artifacts stayNot Produced) - if a live scene already exists, read-only inspection data (
get_objects_summary) provided so real dimensions replace assumptions
Input schema
Required inputs
- user prompt or upstream brief (from
intent-to-3d-brief-writer) - output destination (web/app/game/render)
Optional inputs
- style/mood references
- product constraints and plausible real-world object dimensions
- read-only live-scene summary (
get_objects_summary) if an existing scene is being planned against - performance constraints (e.g., target polycount or mobile web limits)
Assumptions to confirm
- source model availability and scale plausibility (if
get_objects_summaryis unavailable) - quality priority (realtime web vs marketing render)
- acceptable caveat level for deferred sections
Output schema
Primary output
- complete product-hero scene plan covering:
- Subject plan (primary product, context props, plausible real-world dimensions or
get_objects_summarymeasurements) - Composition plan (focal hierarchy, framing intent, negative space)
- Camera plan (angle, lens intent, depth-of-field strategy)
- Lighting plan (key/fill/rim roles, mood intent)
- Material plan (surface finish intent, portability caveats)
- Environment plan (backdrop, support surface, reflections/context)
- Render & export intent (format, resolution/aspect intent, engine direction)
- Web-handoff intent (GLB/realtime readiness notes, or explicit deferral if render-only)
Secondary output
- assumptions register (explicitly listing every unconfirmed input or dimensional assumption)
- targeted clarification questions (when critical inputs are missing or conflicting)
Evidence / caveat output
- explicit deferral log (every omitted or partial plan section must be marked deferred with a concrete caveat)
- fixed truth-state declaration:
Runtime: Not Run,Artifacts: Not Produced(downgrade-only, never upgrade)
Required laws
- ../../laws/official-runtime-only.md
- ../../laws/evidence-before-done.md
- ../../laws/non-blender-user-language.md
- ../../laws/no-arbitrary-python-interface.md
Official runtime boundary
Planning-only skill: no Blender execution, no Python scripts, and no artifact generation.
- Runtime truth state must remain
Not Run(from canonical statesNot Run | Attempted | Produced | Verified | Failed). - Artifact truth state must remain
Not Produced(downgrade-only, never upgrade). - When a live scene exists, consume read-only inspection results (
get_objects_summary) rather than assuming object dimensions, without executing scene mutations.
Operating procedure
- Restate user intent and desired outcome in plain, non-Blender language (or ingest the brief from
intent-to-3d-brief-writer). - Identify target audience and output destination profile (web, app, game, or render).
- Extract hard constraints (brand, style, real-world sizing, delivery platform).
- Generate targeted clarification questions if critical data is missing.
- Build subject plan: define primary hero object, support props, and plausible real-world dimensions (use read-only
get_objects_summarymeasurements instead of assumptions whenever a live scene exists). - Build high-level composition plan (focal hierarchy, framing, negative space) ready for deep-dive in
camera-composition. - Build high-level camera plan (angle, focal length intent, depth strategy) ready for deep-dive in
camera-composition. - Build high-level lighting plan (key/fill/rim roles and mood intent) ready for deep-dive in
lighting-material-plan. - Build high-level material plan (surface finish intent and export portability caveats) ready for deep-dive in
lighting-material-plan. - Build environment plan (backdrop, ground plane, ambient context).
- Define render/export intent and web-handoff intent (ensure every section in steps 5–11 is either present or explicitly deferred with a documented caveat).
- Populate the assumptions register, lock truth states (
Runtime: Not Run,Artifacts: Not Produced), and specify next-skill handoffs (acceptance-criteria-builderfor measurable acceptance rows,done-criteriafor done conditions, andscene-readinessfor post-build review).
Decision tree
- If output destination is unknown → pause and ask for destination first, or provide explicitly caveated options.
- If a live scene exists → use read-only
get_objects_summarydata for object scale/counts instead of guessing; otherwise record plausible real-world dimensions in the assumptions register. - If any plan section cannot be completed → mark that section explicitly deferred with a caveat (never silently omit).
- If constraints conflict (e.g., cinematic glass refractions vs strict mobile web budget) → present Option A / Option B trade-offs and route to
acceptance-criteria-builder. - Regardless of whether downstream runtime is ready, unknown, or blocked → keep planning-only mode with
Runtime: Not RunandArtifacts: Not Produced.
Playbooks
- Playbook A: Text-only planning path (no live scene; relies on plausible real-world sizing assumptions + full assumptions register;
Runtime: Not Run,Artifacts: Not Produced). - Playbook B: Live-scene-informed planning path (consumes read-only
get_objects_summaryto ground object dimensions; prepares handoffs to deep-dive planners;Runtime: Not Run,Artifacts: Not Produced). - Playbook C: Blocked-downstream path (delivers complete scene plan, assumptions register, and blocker caveats while keeping
Runtime: Not RunandArtifacts: Not Produced).
Mode handling
Text-only mode
- deliver complete scene plan and assumptions register using plausible real-world dimensions
- truth states stay
Runtime: Not RunandArtifacts: Not Produced
Runtime-ready mode
- prefer read-only
get_objects_summaryfacts over dimensional assumptions if a live scene exists - include structured handoff steps for downstream planners and checkers
- truth states stay
Runtime: Not RunandArtifacts: Not Produced(no upgrade in this skill)
Blocked runtime mode
- deliver complete scene plan, assumptions register, and blocker-aware next actions
- truth states stay
Runtime: Not RunandArtifacts: Not Produced
Validation checklist
- [ ] intent and output destination restated in plain language
- [ ] subject plan includes plausible real-world dimensions (or
get_objects_summarydata when a live scene exists) - [ ] composition, camera, lighting, material, and environment plans are all present (or explicitly deferred with a caveat)
- [ ] render/export intent and web-handoff intent are present (or explicitly deferred with a caveat)
- [ ] assumptions register is complete and explicit
- [ ] clarification questions included for any unresolved critical inputs
- [ ] truth states explicitly set to
Runtime: Not RunandArtifacts: Not Produced(no overclaims or upgrades) - [ ] no Python code or Blender execution commands included
- [ ] role boundaries respected with explicit handoffs to
camera-composition,lighting-material-plan,acceptance-criteria-builder,done-criteria, andscene-readiness
Pass / Warn / Fail rubric
| Category | Pass | Warn | Fail |
|---|---|---|---|
| Intent & sizing grounding | Intent clear; dimensions grounded via get_objects_summary or plausible real-world sizing | Minor non-critical ambiguity; sizing assumed with caveat | Core intent unclear or implausible/untracked object scale |
| Plan completeness | All 8 plan sections present or explicitly deferred with a caveat | Deferral caveats present but brief | Any plan section silently omitted or left vague without deferral caveat |
| Assumption transparency | Complete assumptions register + clear deferral caveats | Minor assumptions listed inline instead of register | Hidden assumptions or silent guesses |
| Evidence & role boundary | Runtime: Not Run, Artifacts: Not Produced; clean handoffs; no Python | Slightly verbose handoff boundary | Runtime/artifact state upgraded, Python included, or blurs into deep-dive/acceptance/checker roles |
Failure handling
- Missing critical inputs: block and ask targeted questions.
- Conflicting constraints: provide prioritized options with caveats.
- Unknown runtime readiness: continue in text-only planning mode.
Troubleshooting
- Prompt too broad or unstructured: hand off upstream to
intent-to-3d-brief-writerbefore finalizing the scene plan. - Web + cinematic conflict: flag trade-off in assumptions register and route to
cycles-vs-eevee-decisionormobile-glb-budget. - No source model available: log intake risk in assumptions register and route to
asset-discoveryorasset-fallback-strategy.
Best practices
- preserve non-Blender user language and explain visual trade-offs plainly
- keep real-world object dimensions plausible; when a live scene exists, always use read-only inspection (
get_objects_summary) rather than assumptions - ensure every plan section is either complete or explicitly deferred with a caveat
- keep high-level scene planning modular so downstream planners and acceptance builders can consume it cleanly
Good examples
“Complete product-hero plan covering subject (plausible 28 cm shoe length assumed; no live scene get_objects_summary available), composition, camera, lighting, material, environment, and render/export/web-handoff intent, with an explicit assumptions register and truth states locked at Runtime: Not Run and Artifacts: Not Produced.”
Bad examples
“Hero scene done.” (no structure, no caveats)
User-facing response template
- What you asked for: (plain-language goal and destination)
- Proposed product-hero scene plan: (subject & sizing, composition, camera, lighting, material, environment, render/export intent, web-handoff intent)
- Deferred sections & caveats: (any section deferred and why, or "None")
- Assumptions register: (explicit list of unverified inputs and sizing assumptions)
- Current status:
Runtime: Not Run|Artifacts: Not Produced - Next action / handoff: (recommended next skill from catalog)
Anti-patterns
- skipping required laws or runtime boundary statements
- claiming runtime/artifact success without evidence
- claiming runtime execution or upgrading truth states beyond
Runtime: Not RunandArtifacts: Not Produced - assigning
Not Runto artifacts instead of runtime (confusing runtime and artifact truth states) - silently omitting any plan section instead of marking it explicitly deferred with a caveat
- guessing object scale when a live scene exists and read-only
get_objects_summarycan be inspected - blurring roles by writing deep camera/light rigs, measurable acceptance tables, or stop conditions instead of handing off
- including Python code or Blender CLI/MCP execution commands in the plan
- producing Blender-jargon-heavy output without plain-language explanation
Cross-skill handoff
<!-- blendops:generated:start handoffs -->
- [upstream_reference] Upstream input (if raw prompt needs structuring) →
../intent-to-3d-brief-writer/SKILL.md - [progression] Next (deep-dive framing & camera) →
../camera-composition/SKILL.md - [progression] Next (deep-dive lookdev & lighting) →
../lighting-material-plan/SKILL.md - [fallback] Then (measurable pass/fail rows) →
../acceptance-criteria-builder/SKILL.md - [upstream_reference] Then (pin done condition) →
../done-criteria/SKILL.md - [progression] Later (post-build readiness review) →
../scene-readiness/SKILL.md - [conditional] Delivery handoff (when web/render output is targeted) (glb-web-handoff) →
../glb-web-handoff/SKILL.md - [conditional] Delivery handoff (when web/render output is targeted) (render-export-evidence) →
../render-export-evidence/SKILL.md
<!-- blendops:generated:end handoffs -->
Non-goals
- executing Blender commands, MCP mutations, or CLI jobs
- writing Python scripts
- claiming
Attempted,Produced, orVerifiedruntime/artifact states - replacing
intent-to-3d-brief-writer(upstream brief),camera-composition(deep framing),lighting-material-plan(deep lookdev),acceptance-criteria-builder(measurable acceptance rows),done-criteria(done conditions), orscene-readiness(readiness review)
References
- https://www.blender.org/lab/mcp-server/
- https://claude.com/resources/tutorials/using-the-blender-connector-in-claude
- https://docs.blender.org/manual/en/latest/advanced/command_line/index.html