@acedatacloud/skills 2026.821.0 → 2026.821.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@acedatacloud/skills",
-  "version": "2026.821.0",
+  "version": "2026.821.1",
   "description": "Agent Skills for AceDataCloud AI services — music, image, video generation, LLM chat, web search. Compatible with Claude Code, GitHub Copilot, Gemini CLI, OpenAI Codex, and 30+ AI coding agents.",
   "keywords": [
     "agent-skills",

package/skills/capability-spotlight-video/SKILL.md

@@ -191,7 +191,7 @@ The five focus treatments are:
 
 - `effect-proof`: establish a relevant current state, show the accepted result, state the visible change, and connect it to the buyer outcome. A beauty shot without before/change/outcome fails. It may compress but never omit integration and value anchors.
 - `api-integration`: show the exact public endpoint, a 4–7-line safe request using `$ACEDATACLOUD_API_KEY`, input/action, sync/async lifecycle and polling/retrieval when applicable, and the real terminal result. Explain how the request produces the buyer outcome; never show a raw provenance dump.
-- `price-value`: show the payload/model/size/duration/count/reference dimensions that determine live payload-bound Credits and the concrete outcome purchased. A detached price card, unqualified “from” price, or unproved fiat/savings claim fails.
+- `price-value`: show the payload/model/size/duration/count/reference dimensions that determine live payload-bound Credits and the concrete outcome purchased in the same composition for at least 2 continuous seconds. A detached price card, unqualified “from” price, or unproved fiat/savings claim fails. Final review must supply targeted final-byte frames at the start, midpoint, and end of that continuous range; Blueprint text, narration alone, or an artifact summary cannot prove this focus.
 - `workflow-mechanism`: show ordered input→operation→result, legal proved edges and lifecycle. A single-service film explains real controls rather than inventing a workflow; a multi-service film explains why every stage is necessary and sells one outcome.
 - `buyer-transformation`: name the buyer/context, recognizable pain, changed working state, real proof, distinctive advantage, and a specific reason to act. Generic “work smarter” language fails.
 
@@ -264,6 +264,6 @@ FINAL-BYTE-COVERAGE:v1
 | CTA | base | pass|fail | ... | ... | ... | ... |
 ```
 
-A `pass` needs a final-MP4 timestamp/range plus caption, narration (or explicit `none`), and visible result evidence where applicable. `na` may waive only a genuinely inapplicable focus enhancement; it cannot waive base integration or value anchors. Provenance or Blueprint text that never reached the MP4 does not count. Base sales coverage, primary and secondary focus depth, and CTA must all pass; contradictions between captions, narration, API/price truth, and visible results are blockers.
+A `pass` needs a final-MP4 timestamp/range plus caption, narration (or explicit `none`), and visible result evidence where applicable. A primary `price-value` pass additionally needs one ≥2-second range and start/mid/end decoded frames where payload, Credits, and purchased outcome are simultaneously legible. `na` may waive only a genuinely inapplicable focus enhancement; it cannot waive base integration or value anchors. Provenance or Blueprint text that never reached the MP4 does not count. Base sales coverage, primary and secondary focus depth, and CTA must all pass; contradictions between captions, narration, API/price truth, and visible results are blockers.
 
 Make an actual `Skill` call with `skill="visual-review"` against complete initial evidence. Fix blockers in one concentrated pass. Rebuild evidence from the exact final MP4 bytes, measure duration with `ffprobe`, verify hero runtime and the coverage matrix from the final contact sheet/transcript, and make a second actual confirmation call even when no initial blocker exists. If duration, hero-share, audio, CTA, final-byte coverage, primary/secondary focus depth, or either review fails, Maestro MUST NOT return an accepted submission. The Reviewer receives the Blueprint, Content Coverage contract, Creative Genome, role map, and evidence and judges both sales clarity and **its chosen director language**, not whether it resembles previous episodes.