| name | asset-metadata-writer |
| description | Use when writing prompt, asset, review, provenance, score, and decision metadata for generated candidates and promoted assets. |
| version | 1.0.0 |
| author | Project Agents |
| license | Project-internal |
| metadata | {"hermes":{"tags":["metadata","provenance","asset-index"],"related_skills":["asset-gacha-session","score-style-fit","query-asset-candidates"]}} |
Asset Metadata Writer
Overview
Writes the evidence trail for generated assets. No official candidate should exist without metadata.
When to Use
- After generation.
- After scoring/review.
- Before shortlist, scene prep, or candidate index updates.
Required Files
asset.meta.json
prompt.meta.json
review.meta.json
The canonical schema for promoted or reviewable asset metadata is:
content/schema/asset_meta/asset.meta.schema.json
The template under this skill is only a pointer to that canonical schema. Do not introduce a second asset-metadata contract inside skills/.
Candidate Index Row
Append one row to the candidate index when a candidate is created or its status changes. The default project location is:
res://art/experiments/candidate_index.jsonl
Official promotion metadata lives beside the promoted asset:
res://art/official/<content_id>/asset.meta.json
Placeholder assets also require metadata and should use placeholder or PROCEDURAL_PLACEHOLDER status/tag language.
Common Pitfalls
- Saving images without source prompt/seed/tool.
- Replacing metadata instead of preserving review history.
- Not recording rejected candidates that teach useful prompt lessons.
Verification Checklist
