@cloudglue/tinycloud 0.3.0

package/skills/youtube-publish/SKILL.md

@@ -0,0 +1,66 @@
+---
+name: youtube-publish
+description: >-
+  Generate YouTube publishing metadata from a video: title, description,
+  chapter markers, tags, SRT subtitles, and an HTML copy page to paste from.
+  Use when the user is preparing a video for YouTube upload and wants the
+  metadata and subtitles generated. Takes one source: a local video file,
+  URL, or cloudglue:// file URI (e.g. cloudglue://files/<id>). Runs the built-in tinycloud "youtube-publish"
+  workflow; requires the tinycloud CLI configured with a Cloudglue API key
+  (analysis runs through the user's Cloudglue account).
+argument-hint: "[video file, URL, or cloudglue:// file URI]"
+arguments: source
+---
+
+# YouTube publish kit
+
+This skill is a thin wrapper around the `youtube-publish` workflow recipe
+bundled inside the tinycloud binary (`watch → extract → thumbnails →
+captions → render`). It generates SRT subtitles itself as part of the run.
+
+## Run
+
+1. **Check the CLI.** If the general `tinycloud` skill is installed alongside
+   this one, run its `scripts/preflight.sh`. Otherwise verify directly:
+
+   ```bash
+   tinycloud setup --check --json   # ready when data.ok == true
+   ```
+
+   Missing CLI: `npm install -g @cloudglue/tinycloud` (see https://tinycloud.sh).
+   Missing key: `tinycloud setup cloudglue --api-key <key>`.
+
+2. **Confirm the recipe is available** (free, no cloud calls):
+
+   ```bash
+   tinycloud workflow validate youtube-publish --json
+   ```
+
+3. **Run it** with the user's source. The analysis steps run through the
+   configured Cloudglue API key — if the user has not clearly asked to run
+   it, show the step plan first with
+   `tinycloud workflow plan youtube-publish $source --json` (free).
+
+   ```bash
+   tinycloud workflow youtube-publish $source --json
+   ```
+
+   Useful params: `--param segment=chapters` (default; produces the chapter
+   markers YouTube descriptions need) or `--param segment=uniform:20` for a
+   lighter pass on short videos; `--param out=<path>` for the HTML location.
+
+## Read the result
+
+Parse the single JSON envelope from stdout (machine output; logs are stderr):
+
+- Success: `status == "ready"` and `data.status == "completed"`.
+- The copy page is `data.outputs.html`; the run dir also contains the
+  generated SRT under `captions/` (listed in `data.artifacts[]`).
+  Default: `./tinycloud-output/runs/<data.run_id>/youtube-publish.html`.
+- Report the HTML path and the SRT path. The page renders even when
+  subtitles were unavailable.
+
+Any other `status` (`needs_credentials`, `needs_upload`, `pending`, `paused`,
+`error`) or `data.status` of `partial`/`failed`: stop, report the envelope's
+`error.message`, and follow its `setup` / `resume` / `next` hints. The general
+`tinycloud` skill (if installed) documents full status handling.