English | 中文
This is an agent skill suite (9 atomic /command skills) + CLI for Tongji Look (look.tongji.edu.cn):
- login via Tongji IAM SSO (Playwright),
- list courses (recent list or full search),
- transcribe a lecture to
SRT + TXT,
- generate a lecture timeline outline (
*_timeline.txt) from SRT (agent-generated, Simplified Chinese),
- download lecture slide snapshots (filename includes snapshot time),
- then let the current agent write a Markdown study note from transcript + slide images.
Commands
| Command | Description |
|---|
/setup | Configure credentials, check dependencies (Python, Node.js, ffmpeg, vision-support, TeX), set workspace |
/list | List courses, search by keyword, interactive selection |
/trans | Transcribe one lecture to SRT + TXT; optionally download slides in parallel |
/note | Generate study notes + timeline outline from transcript + slides |
/add | Import supplementary materials (PDF, PPTX, DOCX) into a lecture session |
/wiki | Build and locally serve the static course knowledge base |
/page | Deploy the built course wiki to GitHub Pages via gh CLI |
/cheatsheet | Generate A4 cheat sheet (LaTeX or HTML) from course notes |
/ralphtrans | Batch transcribe all lectures in a course with checkpoint/resume |
Install
Recommended: npx skills
If your agent supports the skills protocol:
npx skills install https://github.com/walkerkiller/look-tongji-notes
Claude Code Marketplace
In Claude Code:
/plugin marketplace add https://github.com/walkerkiller/look-tongji-notes
/plugin install look-tongji-notes
Other Platforms
| Platform | How |
|---|
| Claude Code | Marketplace (above) or point plugin root to this repo |
| Codex CLI | .codex-plugin/plugin.json → ./skills/ |
| Cursor | .cursor-plugin/plugin.json → ./skills/ |
| Gemini CLI / OpenClaw / OpenCode / Hermes Agent | plugin.json → ./skills/ |
Usage (CLI)
<SKILL_DIR> is the folder that contains SKILL.md.
Setup credentials (recommended):
python "<SKILL_DIR>/../../scripts/look_tongji.py" setup
List recent courses:
python "<SKILL_DIR>/../../scripts/look_tongji.py" list
Search courses by name (recommended for accuracy, calls get_all_courses internally):
python "<SKILL_DIR>/../../scripts/look_tongji.py" list --all --query "<COURSE_NAME_KEYWORD>"
Transcript only (transcribe, aliases transcript / trans):
python "<SKILL_DIR>/../../scripts/look_tongji.py" transcribe --lecture-url "<LECTURE_URL>"
Combined mode (note, runs transcript + slide in parallel by default):
python "<SKILL_DIR>/../../scripts/look_tongji.py" note --lecture-url "<LECTURE_URL>"
Note style (affects how the generated note is formatted):
python "<SKILL_DIR>/../../scripts/look_tongji.py" note --lecture-url "<LECTURE_URL>" --note-style dialogue
Supports standard (lecture notes, default) and dialogue (Q&A format).
[!TIP]
The CLI detects lectures shorter than 1 hour and prints a non-blocking warning: [Warning] 课时不足1小时 — suggesting a retry, as very short lectures may indicate an incomplete recording or playback error.
Download slide snapshots for a lecture:
python "<SKILL_DIR>/../../scripts/look_tongji.py" slide --lecture-url "<LECTURE_URL>"
If throttling is suspected, reduce concurrency:
python "<SKILL_DIR>/../../scripts/look_tongji.py" slide --course-id "<COURSE_ID>" --sub-id "<SUB_ID>" --concurrency 2 --retries 5
In the /note workflow, the agent generates a timeline outline after the SRT subtitle file is produced:
- File:
./tongji-output/<course_id>_<sub_id>_timeline.txt
- One line per segment (Simplified Chinese), format:
Start-Over：Stage Main Content
- Example:
00:00-05:30：Course Orientation and Assessment Description
- Skip only if the user explicitly says:
no outline / no timeline.
Artifacts are written to the configured course-wiki workspace by default.
Agent Note
When a user says /setup / /list / /trans / /note / /wiki / /add / /page / /cheatsheet / /ralphtrans, follow the corresponding skills/<name>/SKILL.md and run the matching CLI commands in scripts/look_tongji.py.
