| name | render-export-evidence |
| description | Verify preview/render/GLB artifact truth by recording commands, inputs, output paths, file existence, validation notes, limitations, and conservative status labels. |
render-export-evidence
Purpose
Create a conservative evidence ledger for preview, render, export, and GLB claims. Use this skill to decide whether an artifact is Not Run, Attempted, Produced, Verified, or Failed.
This skill verifies evidence; it does not run Blender or create artifacts by itself.
Quick start
- Identify the claimed artifact or runtime action.
- Collect command/tool, input/script/scene, output path, file existence, validation notes, and limitations.
- Assign the most conservative valid status.
- Hand off to
glb-web-handoff or response writer only after evidence is clear.
When to use
Use this skill when:
- a user asks whether preview/render/GLB/export succeeded
- an agent needs to summarize artifact evidence
- a runtime eval produced or attempted files
- a report needs artifact truth labels
- a handoff needs proof before user-facing claims
When not to use
Do not use this skill when:
- the user only needs a scene plan with no runtime/artifact evidence
- runtime readiness has not been checked and the user wants execution
- the task is GLB web integration language after evidence is already summarized; use
glb-web-handoff
- the request asks to run Blender, render, or export; use runtime/eval flow first
Trigger phrases
- “Did the render produce?”
- “Can I claim the GLB exists?”
- “Verify export evidence.”
- “What artifact status should this have?”
- “Summarize preview/render/GLB evidence.”
Prerequisites / readiness
- The claimed artifact type is known: preview, render, Blender file, GLB/export, or handoff record.
- Any runtime action is already completed or explicitly
Not Run / Attempted.
- Evidence sources are available or can be marked missing.
- No artifact claim is accepted from intent, prompt text, or transcript alone.
Input schema
Required inputs
| Input | Why it is required |
|---|
| Artifact type | Status is scoped per artifact. |
| Runtime path or mode | Separates Path 1 (with host a or b), Path 2, CLI fallback, or text-only mode (see ../../docs/runtime-stack-strategy.md). |
| Command/tool/action used | Proves how the artifact was attempted. |
| Input/script/scene | Connects output to a specific run. |
| Output path | Required for file-based claims. |
| File existence or visible evidence | Required before Produced or Verified. |
| Validation notes | Required before Verified. |
| Limitations/failures | Prevents overclaiming. |
Optional inputs
| Input | Use |
|---|
| File size/hash | Strengthens generated-file evidence. |
| Screenshot/render thumbnail path | Supports visible-output claims. |
| Logs/exit status | Supports CLI or failed-attempt analysis. |
| User-facing target | Shapes handoff language. |
Assumptions to confirm
- Runtime eval remains
Not Run unless an actual runtime action was recorded.
- Preview/render/GLB remains
Not Produced unless concrete output evidence exists.
Produced is not Verified until validation notes exist.
Output schema
Primary output
An artifact evidence ledger.
Secondary output
- status per artifact type
- missing evidence list
- next safe action
- handoff recommendation
Evidence / caveat output
Artifact type: <preview | render | blend | GLB/export | handoff>
Status: Not Run | Attempted | Produced | Verified | Failed
Runtime path: <Path 1 host a (Anthropic Connector) | Path 1 host b (manual MCP) | Path 2 (ahujasid) | CLI fallback | text-only | unknown>
Blender version: <output of `blender --version` or unknown>
Blender-side add-on identity: <Lab MCP add-on version | ahujasid addon.py commit | N/A for CLI>
MCP server source: <Lab .mcpb bundle filename or commit | uvx blender-mcp version | N/A for CLI>
MCP host product: <Claude Desktop (Connector or manual MCP) | Cursor | Codex | OpenCode | Cline | other | N/A for CLI>
Command/tool/action: <recorded or missing>
Input/script/scene: <recorded or missing>
Output path: <path or Not Produced>
File existence: <yes/no/unknown>
Validation notes: <notes or missing>
Limitations: <caveats>
Next safe action: <action>
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
This skill can evaluate evidence from the 2-path + CLI appendix runtime model (see ../../docs/runtime-stack-strategy.md):
- Path 1 — Official Blender Lab MCP (Lab add-on + Lab server in Blender 5.1+, hosted from either (a) Anthropic Blender Connector in Claude Desktop, or (b) any other MCP client configured manually). Both hosts share the same Blender-side stack and tool surface.
- Path 2 — Community
ahujasid/blender-mcp (different addon.py + server via uvx blender-mcp, mature 21K+ stars third-party, Blender 3.0+).
- CLI fallback (appendix) — direct
blender --background --python, no MCP. Documented upstream as a first-class Blender CLI surface (stable across LTS releases); no in-repo evidence file yet.
Evidence records must name the path explicitly (and Path 1 host option a/b when relevant) and must include Blender version + Blender-side add-on identity + MCP server source/commit + MCP host product so the record is unambiguously attributable.
This skill must not run Blender, render, export GLB, mutate scenes, install runtimes, or promote optional local bridge evidence as official release-eval evidence.
Operating procedure
- Name every artifact type in scope.
- For each artifact, collect:
- command/tool/action used
- input/script/scene
- output path
- file existence or visible-output evidence
- validation notes
- limitations/failures
- Assign status using the decision tree.
- List missing evidence before any handoff.
- Recommend next safe action.
Decision tree
No runtime action recorded?
→ Not Run
Runtime action recorded but output missing/incomplete?
→ Attempted or Failed
Output path/visible evidence exists but validation missing?
→ Produced
Output exists and validation notes pass?
→ Verified
Runtime/action failed or artifact absent when required?
→ Failed or Not Produced with reason
Playbooks
Playbook A: Text-only planning output
- Runtime status:
Not Run.
- Artifact status:
Not Produced.
- Next action: runtime readiness or eval packet.
Playbook B: CLI fallback output
Require:
- exact Blender command
- script/input path
- output folder
- generated file paths
- exit status/logs
- validation notes
Without generated files, do not claim artifacts.
Playbook C: Connector output
Require:
- connector action/tool name or UI action
- Blender-side action notes
- generated file paths or screenshot/render evidence
- validation notes
- permission/session caveats
Read-only smoke evidence is not render/export evidence.
Playbook D: Path 2 (community ahujasid/blender-mcp) output
- Label as Path 2 with upstream commit/version of
addon.py and uvx blender-mcp.
- Keep separate from Path 1 evidence — Path 2 is third-party from both Anthropic and the Blender Foundation.
- Record host/port and arbitrary code execution caveats (
execute_blender_code runs LLM-generated Python with no sandbox).
- See ../../docs/unofficial-runtime-bridges.md for the full Path 2 caveat write-up.
Mode handling
Text-only mode
Use to mark artifacts Not Run / Not Produced and prepare evidence checklist.
Runtime-ready mode
Use after the runtime/eval action has happened or when reviewing a captured eval record.
Blocked runtime mode
Use when runtime was unavailable; record blockers and keep artifacts Not Produced.
Validation checklist
Pass / Warn / Fail rubric
| Verdict | Criteria |
|---|
| Pass | Status labels match evidence, output paths/validation are clear, and limitations are explicit. |
| Warn | Some evidence exists but validation, provenance, or scope is incomplete. |
| Fail | Artifact success is claimed without evidence, or labels contradict the recorded facts. |
Failure handling
- Missing output path → status cannot exceed
Attempted.
- Missing file existence/visible evidence → status cannot exceed
Attempted.
- Missing validation notes → status cannot exceed
Produced.
- Failed command/export → record
Failed or Not Produced with reason.
- Pre-existing file only → do not claim current run produced it unless provenance links it to the run.
Troubleshooting
| Problem | Conservative handling |
|---|
| Screenshot exists but no render file | Mark preview evidence separately from render file. |
| GLB file exists but no validation | Produced, not Verified. |
| CLI log says success but file missing | Attempted or Failed, not Produced. |
| Output path unknown | Attempted; request path evidence. |
| Read-only connector smoke passed | Runtime access scoped to read-only; artifacts Not Produced. |
Best practices
- Track each artifact type separately.
- Prefer paths and validation notes over prose claims.
- Use exact status words.
- Keep user-facing summary understandable for non-Blender users.
- Hand off to
glb-web-handoff only after GLB/export status is known.
Good examples
- “Render status: Produced. Output path:
out/hero.png. Validation: camera framing checked; material fidelity not yet verified.”
- “GLB status: Not Produced. Runtime action was Not Run; no export path exists.”
Bad examples
- “GLB produced” with no file path.
- “Render verified” with no validation notes.
- “Runtime eval passed” from read-only smoke evidence only.
User-facing response template
Artifact evidence summary
- Runtime stack/mode: <state>
- Preview: <status + path/evidence>
- Render: <status + path/evidence>
- GLB/export: <status + path/evidence>
- Validation: <notes or missing>
- Limitations: <caveats>
- Next safe action: <action>
Anti-patterns
- Treating a planned output as a produced artifact.
- Treating a prompt transcript as file evidence.
- Merging preview, render, and GLB into one vague status.
- Calling optional local bridge evidence official release evidence.
Cross-skill handoff
- Runtime readiness →
../official-runtime-readiness-checker/SKILL.md
- Scene validation →
../blender-scene-quality-checker/SKILL.md
- GLB/web handoff →
../glb-web-handoff/SKILL.md
- User-facing explanation →
../non-blender-user-response-writer/SKILL.md
Non-goals
- Run Blender.
- Render images.
- Export GLB.
- Modify scene data.
- Install runtime bridges.
- Publish artifacts.
References
../../docs/skill-system.md../../docs/runtime-stack-strategy.md../../docs/evals/official-runtime-verification-criteria.md../../docs/evals/official-runtime-manual-eval-packet.md../../laws/evidence-before-done.md
