一键导入
compare-video-runs
Compare two local video artifact bundles and surface report-status changes, artifact presence changes, and coarse video-level deltas.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Compare two local video artifact bundles and surface report-status changes, artifact presence changes, and coarse video-level deltas.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Assemble a grounded review prompt from a local artifact bundle so Codex or Claude Code can inspect the right files and ask the right questions.
Check a local video against layout annotations, safe zones, OCR text boxes, and frame samples so agents can catch cropped content, caption overlap, unreadable overlays, and platform-unsafe composition before shipping.
Collect first-pass source media facts, audio silence/energy, shot estimates, representative frame status, and persistent text/caption risk evidence.
Review a rendered video for technical defects such as resolution mismatch, missing or near-silent audio, black or white frames, edge gutters, low motion, sparse caption bands, and layout issue pass-through.
Fuse video shot segments with storyboard frames, OCR, transitions, and timeline artifacts into one per-segment evidence map for grounded video review.
Extract at least one storyboard frame per video shot segment so later OCR and segment evidence fusion can cover gaps missed by global sampling.
| name | compare-video-runs |
| description | Compare two local video artifact bundles and surface report-status changes, artifact presence changes, and coarse video-level deltas. |
Use this when an agent needs a quick regression view across two output bundles.
This is best for:
Provide two bundle locators using the same shape accepted by
video-artifact-intake:
{
"left": {
"outputDir": "./output/baseline-run"
},
"right": {
"outputDir": "./output/current-run"
}
}
Use left for the baseline or expected-good run and right for the candidate
run.
The result compares normalized bundle maps: overall status changes, artifact presence changes, report status changes, and coarse video metadata deltas when both sides can be probed.
Run video-artifact-intake first if either side is ambiguous. Use
review-bundle on the candidate side when the comparison finds warnings or
failures. Use storyboard skills on both sides only when visual regressions need
evidence beyond report/status deltas.
The comparison is only as strong as the artifacts present in both bundles. A missing report on the right may indicate a regression in generation, a changed pipeline, or a different output directory. Video duration/resolution deltas are coarse signals; inspect the videos or storyboard frames for behavioral changes.
Do not call a run better solely because it has fewer reports or an unknown
status. Abstain from visual regression conclusions unless both runs have
comparable video/storyboard evidence.
If left and right point to the same directory, the comparison will be
meaningless. Stale latest pointers can compare the wrong run. Missing ffprobe
removes video metadata deltas but should not block artifact comparison.
Repo-side runner:
node --import tsx scripts/harness/compare-bundles.ts
Installed-pack runner:
node ./.video-evaluator/agent/run-tool.mjs compare-video-runs