From 6631e4146e5ec1a8c26bfe818428f31e0b569d90 Mon Sep 17 00:00:00 2001 From: Jelmer de Wit <1598297+jdwit@users.noreply.github.com> Date: Wed, 1 Jul 2026 20:41:31 +0200 Subject: [PATCH 1/2] docs: document agentic authoring workflow in README (#43) --- README.md | 30 +++++++++++++++++++++++++++--- 1 file changed, 27 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index f9d5e02..02f7abf 100644 --- a/README.md +++ b/README.md @@ -30,20 +30,44 @@ command line. - Bulk update video metadata across hundreds of videos in one pass (search-replace titles, descriptions, tags). -- Supply the inputs for transcript-grounded, on-brand metadata authoring: read caption tracks, pull - a transcript as clean text, and store a per-channel brand voice, so an agent (or you) can write - the copy and apply it with `videos update`. - Upload videos from a directory using YAML sidecars for metadata and thumbnails. - Schedule, start, stop, and update YouTube livestream broadcasts, including RTMP ingest details. - Multi-channel profiles: manage several channels from one machine and switch per command. - Comments moderation: list, reply, approve, reject, and ban from the CLI. - Channel analytics queries via the YouTube Analytics API. - Playlists: bulk-add by search and reorder by views with one command. +- [Agentic authoring](#agentic-authoring): edit descriptions and metadata with the help of an AI + agent, grounded in the video's actual transcript and shaped by a per-channel brand voice. ## Documentation See the [full documentation](https://jdwit.github.io/ytstudio-cli/) for installation, OAuth setup, and the command reference. +## Agentic authoring + +Writing a good title or description is a creative, per-video task. Instead of adding an +"AI" command, this CLI gives an agent the raw materials to do it well and stay honest, +then applies the result through the ordinary update path: + +- **Transcript grounding.** `videos captions ` lists a video's caption tracks and + `videos transcript [--lang nl]` pulls one as clean plain text (timestamps and + SRT markup stripped), so the copy is based on what was actually said rather than a + hallucination. +- **Per-channel brand voice.** `profile brand edit` (opens `$EDITOR`) or + `profile brand set --file ` store a free-form `brand.md` next to each profile, + holding the house style (audience, tone, title conventions) an agent should follow so + output sounds on-brand instead of generic. +- **Apply it.** The agent drafts the metadata and applies it through the existing + `videos update --title ... --description ...` path (dry-run by default, then + `--execute`). + +The CLI itself never calls a model: it supplies the inputs and applies the change, +while a skill-aware AI agent (such as [Claude Code](https://claude.com/claude-code) or Pi) +writes the copy by running the [agent skill](skills/ytstudio/SKILL.md) below. The transcript is the source of truth for +claims and the brand file is the source of truth for tone, so the agent never states +anything the transcript does not support. There is no LLM or API key to configure here; +that lives with whichever agent you run. + ## Agent skill This repo includes a harness-agnostic [agent skill](skills/ytstudio/SKILL.md) that From d50fbb72ca98830fdfa182a0ffc547f1ac7376ba Mon Sep 17 00:00:00 2001 From: Jelmer de Wit <1598297+jdwit@users.noreply.github.com> Date: Wed, 1 Jul 2026 20:46:20 +0200 Subject: [PATCH 2/2] docs: condense agentic authoring to a single feature bullet --- README.md | 29 ++--------------------------- 1 file changed, 2 insertions(+), 27 deletions(-) diff --git a/README.md b/README.md index 02f7abf..65de8eb 100644 --- a/README.md +++ b/README.md @@ -36,38 +36,13 @@ command line. - Comments moderation: list, reply, approve, reject, and ban from the CLI. - Channel analytics queries via the YouTube Analytics API. - Playlists: bulk-add by search and reorder by views with one command. -- [Agentic authoring](#agentic-authoring): edit descriptions and metadata with the help of an AI - agent, grounded in the video's actual transcript and shaped by a per-channel brand voice. +- Agentic authoring: use your LLM to write descriptions and edit metadata, based on the actual + video transcript and guided by a per-channel tone of voice. ## Documentation See the [full documentation](https://jdwit.github.io/ytstudio-cli/) for installation, OAuth setup, and the command reference. -## Agentic authoring - -Writing a good title or description is a creative, per-video task. Instead of adding an -"AI" command, this CLI gives an agent the raw materials to do it well and stay honest, -then applies the result through the ordinary update path: - -- **Transcript grounding.** `videos captions ` lists a video's caption tracks and - `videos transcript [--lang nl]` pulls one as clean plain text (timestamps and - SRT markup stripped), so the copy is based on what was actually said rather than a - hallucination. -- **Per-channel brand voice.** `profile brand edit` (opens `$EDITOR`) or - `profile brand set --file ` store a free-form `brand.md` next to each profile, - holding the house style (audience, tone, title conventions) an agent should follow so - output sounds on-brand instead of generic. -- **Apply it.** The agent drafts the metadata and applies it through the existing - `videos update --title ... --description ...` path (dry-run by default, then - `--execute`). - -The CLI itself never calls a model: it supplies the inputs and applies the change, -while a skill-aware AI agent (such as [Claude Code](https://claude.com/claude-code) or Pi) -writes the copy by running the [agent skill](skills/ytstudio/SKILL.md) below. The transcript is the source of truth for -claims and the brand file is the source of truth for tone, so the agent never states -anything the transcript does not support. There is no LLM or API key to configure here; -that lives with whichever agent you run. - ## Agent skill This repo includes a harness-agnostic [agent skill](skills/ytstudio/SKILL.md) that