feat(skills): ai-presenter-video optional skill (port of lanshu, 955★ MIT)
Ports cclank/lanshu-create-ai-presenter-video (MIT, 955 stars in 7 days)
into optional-skills/creative/ai-presenter-video. Provider-neutral
presenter-video production: locked narration as master clock, avatar
generation with pilot-first cost discipline, lip-sync/identity QA,
captions, deterministic ffmpeg finalization with loudness normalization
and contact-sheet verification.
Hermes adaptations in the hub SKILL.md: SKILL_DIR resolution (upstream
hardcoded ~/.codex/skills), capability mapping to text_to_speech / FAL
video families / vision_analyze / hyperframes, consent-flag JSON paths
(input.* vs root), preflight error-vs-remote-blocker semantics.
References kept substantively verbatim (all-English upstream). Scripts
unmodified. LICENSE carried.
Validated hands-on: init_job -> preflight gating (blocked until manual
review booleans + input.remote_upload_approved) -> finalize_delivery on
a synthetic 1080x1920 render (master+share decode-verified, delivery
report, 9-frame contact sheet). Cold-subagent live test: SHIP; 3
friction fixes folded in (resolution guard, boolean-flip example,
preflight-writes-job note).
Teknium
a79ff58d65