Asset fallback strategy

asset-fallback-strategy

Purpose

When asset-license-checker / asset-style-consistency-checker / glb-mobile-performance-budget blocks an asset, this skill produces an actionable fallback (replace / re-source / generate procedural / drop with caveat) so the recipe can proceed.

Quick start

• read the failed asset + reason
• propose 2-3 fallback options ordered by speed-to-implement
• pick one with user agreement
• update the asset list

When to use

• after a checker fails an asset
• when an asset cannot be acquired in time
• when budget excludes the originally-planned asset
• when license blocks an asset

When not to use

• when no asset has failed (nothing to fall back to)
• pure exploration (no concrete asset list)
• when the user wants to wait for the original asset

Trigger phrases

• "what if we can't get this asset"
• "fallback for the model"
• "the license blocks us"
• "scaled-down asset"

Prerequisites / readiness

• failed asset + reason known
• recipe brief known (informs alternative selection)

Input schema

Required inputs

Input	Why it is required
Failed asset + reason	Drives fallback search
Recipe brief	Alternative must satisfy brief
Time / budget remaining	Affects feasibility of options

Optional inputs

Input	Use
Existing internal asset library	Source of fast fallbacks
Generative tool access (procedural / AI)	Adds option types

Assumptions to confirm

• The user accepts that a fallback is needed (vs waiting).
• "Drop with caveat" is a valid option even if not preferred.

Output schema

Primary output

A fallback option table: option (replace / re-source / procedural / generative / drop), source / method, time-to-implement, brief match, license confidence, recommendation.

Secondary output

• updated asset list reflecting picked fallback
• caveat lines if "drop" is chosen
• evidence label

Evidence / caveat output

Runtime status: Not Run | Attempted | Produced | Verified | Failed | Blocked / Not Run
Artifact status: Not Run | Not Produced | Produced | Verified | Failed
Evidence used: <links, paths, logs, or "none">
Limitations: <known gaps>

Required laws

• ../../laws/evidence-before-done.md
• ../../laws/non-blender-user-language.md
• ../../laws/no-arbitrary-python-interface.md
• ../../laws/official-runtime-only.md

Official runtime boundary

Process / planning skill — does not download / acquire / generate assets. Outputs a strategy.

Operating procedure

1. Read failed asset + reason.
2. Read brief style + budget context.
3. Generate 2-3 fallback options (see references/fallback-option-types.md).
4. Score each on time / brief match / license / quality.
5. Recommend one; require user agreement.
6. Update asset list; flag downstream rework needed.

Decision tree

Failed by license?
  → Replace from CC0 source first; otherwise paid alternative
Failed by budget (size/poly)?
  → Lower-poly version; or procedural; or simplification
Failed by style?
  → Replace from same-style source; or restyle
Failed by unavailability?
  → Re-source from cache; or procedural; or drop with caveat

Playbooks

Playbook A: Hero shoe model unavailable

Options: (1) Use stand-in mesh from internal lib (fastest, may not match brand). (2) Source from photogrammetry service (slowest, exact). (3) Procedural / generative (medium, brand-aligned but stylization may drift).

Playbook B: HDRI blocked by license

Options: (1) Replace with CC0 from polyhaven (fastest). (2) Bake HDR from procedural sky shader (medium). (3) Drop HDRI; rely on explicit lights (caveat: ambient feel changes).

Playbook C: Texture over-budget

Options: (1) Resize from 4K to 1K (fastest). (2) Replace with procedural noise + tiled detail (medium). (3) Drop texture variation; use solid color (caveat: visual difference).

Mode handling

Text-only mode

Strategy table only.

Runtime-ready mode

After picking fallback, hand off to discovery / asset-fetch action.

Blocked runtime mode

Strategy is forward-plan; no immediate action.

Validation checklist

• Failed asset + reason explicit
• 2-3 options with time / match / license columns
• Recommendation explicit
• User agreement evidence required
• Updated asset list reflected
• Caveat lines if "drop"

Pass / Warn / Fail rubric

Verdict	Criteria
Pass	2-3 options + recommendation + agreement + updated list.
Warn	Options present but recommendation soft.
Fail	"Just drop it" without caveat; no alternatives explored; agreement missing.

Failure handling

• No viable fallback exists → recommend recipe scope reduction (drop the deliverable that needed this asset).
• Fallback also blocks → repeat the cycle with new fallback search.
• User refuses all options → document; flag as recipe-blocker.

Troubleshooting

Problem	Response
Fallback doesn't match brief	Either change brief or change fallback; do not silently lower quality.
Procedural option requires modeling skill not available	Cross out; rely on replace / drop.
Generative AI tool license unclear	Run through asset-license-checker first.

Best practices

• Always offer at least 2 options.
• Always show a "drop with caveat" option (even if last choice).
• Time-to-implement is decisive; brief match is necessary; license confidence is non-negotiable.
• Keep this skill narrow.

Good examples

• "Failed: hero-shoe-mesh.glb (license-blocked). Options: (1) polyhaven CC0 generic shoe — 30 min, brief match low. (2) Buy commercial license $99 — 2 hr, exact match. (3) Procedural shape + texture — 4 hr, match medium. Recommended: (2) for brief fidelity. Decision needed."

Bad examples

• "We'll figure it out." — no options listed.

User-facing response template

Failed asset: <name>
Reason: <license / budget / style / unavailable>

| Option | Source / method | Time | Brief match | License | Recommendation |
|---|---|---|---|---|---|

Recommendation: <option>
Decision needed before continuing.
Caveat lines (if drop): <list>
Updated asset list: <link or note>
Limitations: <gaps>
Next: discovery for selected fallback / scope reduction if all blocked

Anti-patterns

• "Just drop it" without caveat.
• Silently substituting without user agreement.
• Inventing fallback sources.
• Skipping the license check on the new fallback.

Cross-skill handoff

• Discovery for new asset → ../blender-asset-discovery-planner/SKILL.md
• License check on fallback → ../asset-license-checker/SKILL.md
• Style check on fallback → ../asset-style-consistency-checker/SKILL.md
• Budget recheck → ../glb-mobile-performance-budget/SKILL.md
• Library lookup → ../asset-library-organization-planner/SKILL.md

Non-goals

• Acquire the asset.
• Run Blender.
• Generate AI imagery.

References

• references/fallback-option-types.md
• references/scoring-rubric.md
• references/drop-with-caveat-rules.md
• ../../docs/skill-system.md