daily-workflow

Orchestrates the full content intelligence pipeline. Maximises throughput by firing slow async jobs first, then handling fast sync tasks, then collecting async results.

Prerequisites: gws CLI, agent-browser, yt2doc. Refer to ../gws-shared/SKILL.md for auth.

Procedure

Step 1 — Discover and classify Delegate tasks

gws tasks tasklists list

Find the item where title == "Delegate". Extract its id. If not found, skip Steps 2–5.

gws tasks tasks list \
  --params '{"tasklist": "<DELEGATE_LIST_ID>", "showCompleted": false, "maxResults": 100}'

For each task, scan title, links[].link, links[].description, notes for a URL:

Step 2 — Fire YouTube background jobs

For each task in youtube_queue:

📄 Read ../ingest-youtube/SKILL.md §Step 1 (Video Strategist) to determine model

mkdir -p reports/YouTube_YYYY_MM_DD

yt2doc \
  --video "<YouTube URL>" \
  --output ./reports/YouTube_YYYY_MM_DD/<video_id>.md \
  --whisper-model <model> \
  --add-table-of-contents

Use run_command with WaitMsBeforeAsync=5000. Store: { task_id, youtube_url, command_id, output_path, model }.

Tell the user what's happening: "Launching yt2doc for <url> (~X–Y min). Running in background while I process other tasks."

Step 3 — Process newsletters

📄 Follow ../ingest-newsletter/SKILL.md

(Full lifecycle: fetch → summarise → write report → append suggestion → mark read)

Step 4 — Process Threads tasks

For each task in threads_queue:

📄 Follow ../ingest-threads/SKILL.md

(Full lifecycle: fetch → summarise → write report → append suggestion → mark task done)

Step 4W — Process Website tasks

For each task in website_queue:

📄 Follow ../ingest-website/SKILL.md

(Full lifecycle: fetch via Jina Reader → summarise → write report → append suggestion → mark task done)

Log soft failures: "⚠️ Skipping '<title>': fetch failed (Jina + content-cleaner both failed)."

Step 5 — Complete YouTube tasks

For each YouTube background job (from Step 2):

1. Poll: command_status(command_id, WaitDurationSeconds=60) — repeat until DONE. Report elapsed time periodically.
2. On exit code != 0, handle per ../ingest-youtube/SKILL.md §Step 3 error handling. Do NOT mark task done.
3. Read transcript file.
4. Generate summary:

📄 Read ../content-summary/references/summarise.md

1. Write report:

📄 Read ../content-summary/references/filename_rules.md 📄 Read ../content-summary/references/output_template.md

1. Append suggestion:

📄 Read ../content-summary/references/ai_analysis.md 📄 Follow ../content-summary/references/suggestion_log.md

({SourceType} = YouTube)

1. Mark task done + cleanup:

   gws tasks tasks patch \
     --params '{"tasklist": "<DELEGATE_LIST_ID>", "task": "<TASK_ID>"}' \
     --json '{"status": "completed"}'
   rm "reports/YouTube_YYYY_MM_DD/<video_id>.md"

Step 6 — Distill

📄 Follow ../daily-distiller/SKILL.md

Step 7 — Review suggestions

📄 Follow ../review-suggestions/SKILL.md

Final Summary

Processed N newsletter(s) → reports/Newsletter_YYYY_MM_DD/
Processed T Threads task(s):
  ✅ @handle — topic → reports/Threads_YYYY_MM_DD/filename.md
Processed W Website task(s):
  ✅ example.com — Article Title → reports/Website_YYYY_MM_DD/filename.md
  ⚠️ another.com — Article Title → FAILED (Jina + content-cleaner both failed)
Processed Y YouTube task(s):
  ✅ Video Title → reports/YouTube_YYYY_MM_DD/filename.md
  ⚠️ Another Video → FAILED (OOM — increase system RAM)
Skipped Z task(s) (no URL found).
Distillation complete. Suggestions reviewed.