Video Multiplier

Turn one YouTube upload into the full package: chapters, five title options and the description, then up to six post drafts built from the transcript (an X thread, a LinkedIn post, a newsletter section and two or three short-form scripts). It writes text and never cuts clips. Publishing stays with the user.

Step 0 — tools

agnt_tools_list_enabled; if data:youtube or database families are missing, agnt_tools_enable({ "families": ["data:youtube", "database"] }). A requires_upgrade result carries a confirm_url — give it to the user and wait, then re-run. (Connecting through the video-multiplier skill endpoint pre-enables these.)

Tool names on grouped servers

This skill writes flat (per-operation) tool names. New superagnt servers list grouped tools instead: one tool per resource with an action argument. Flat names still resolve there but are not listed, so call the grouped form:

• data_youtube_youtube_channel_id / data_youtube_channel_videos → data_youtube_channels, action channel_id / videos
• data_youtube_video_details / data_youtube_video_subtitles → data_youtube_videos, action details / subtitles
• agnt_db_insert / agnt_db_upsert → agnt_db_write, action insert / upsert
• agnt_db_apply_migration → agnt_db_sql, action apply_migration
• agnt_agents_create / agnt_agents_deploy → agnt_agents_write, action create / deploy
• agnt_schedules_create → agnt_schedules_write, action create
• agnt_tools_enable → agnt_tools_write, action enable

On a grouped server every data source is already on, so skip the data:* part of Step 0 (enable any other family with agnt_tools_write, action enable). Grouped data calls return a compact markdown view: pass response_format: "json" for every view field, or "raw" for the full payload, when a step needs a field the table leaves out.

Step 1 — intake

Take the video link directly, or resolve the user's channel (data_youtube_youtube_channel_id, data_youtube_channel_videos) so new uploads can queue themselves. On first run, ask what they plug in every post (links, products, sponsor lines) and where their audience lives (X, LinkedIn, shorts, a newsletter) — that decides which derivative drafts to produce. Store it:

• multiplier_profile(id, standing_links jsonb, channels text[], voice_notes text, updated_at timestamptz) — via agnt_db_apply_migration then agnt_db_insert.
• multiplier_packages(video_id primary key, url, title, status text default 'queued', chapters_md text, title_options jsonb, description_md text, packaged_at timestamptz).
• multiplier_posts(id, video_id, kind text, draft_md text, source_ts text, status text default 'draft', created_at timestamptz) — kind ∈ thread / short_script / linkedin / newsletter_section.

Step 2 — package the upload

data_youtube_video_details for title, stats and length, then data_youtube_video_subtitles for the transcript. From the transcript write the studio package: chapters with mm:ss timestamps at the real topic turns, five title options (the honest ones — no bait the video does not pay off), and the full description with the standing links. Save on multiplier_packages (agnt_db_upsert) and show it ready to paste. No subtitles → say so and stop; never package from the title alone.

Rank the five titles by the pattern each one uses, and show the rank. In superagnt's study of 1,412 long-form videos from 15 AI and builder channels (pulled 2026-09-30; a hit is 3x the channel's own median views, 13% of all videos), the share of videos with each title pattern that hit:

A pattern earns a title only when the video delivers it: "full course" for a course, "clearly explained" for an explainer. The buckets overlap and several are small, so treat the rank as a tiebreaker, not a rule.

Step 3 — draft the posts

From the same transcript, draft the posts for the channels the user named: an X thread built on the strongest sequence, 2-3 short-form scripts each anchored on one moment (hook, beats, CTA), a LinkedIn post, a newsletter section. Every draft quotes or paraphrases the video's actual lines and carries its source timestamp in source_ts. File each into multiplier_posts and present the set for review. The user edits and posts; nothing goes out by itself.

Step 4 — offer the per-upload run (a task agent)

Once one package has shipped, offer to make it automatic per upload by building a task agent — the full recipe is the base task-agents skill. Short form: agnt_agents_create with a system prompt carrying steps 2–3 verbatim (tables, package shape, the channel list), add ONLY the tools those steps use, agnt_agents_deploy, then agnt_schedules_create on a daily check of the channel for new uploads — cadence confirmed with the user first; every firing is a billed run. The lifecycle and schedules families are on every plan, and plans cap agents and schedules; requires_upgrade → confirm_url → wait. If the user prefers local-only, skip this and tell them to re-run the skill per upload.

Rules

• Every chapter, quote and claim comes from the transcript or the video details — never from memory of the topic.
• Titles never promise what the video does not deliver.
• This skill drafts; it never uploads, posts, comments, or edits the user's YouTube studio.