x-extractor (pure Python skill, Claude Code only)

All the work is done by python_skills/confirmed/x_pipeline/x_link.py. There is no LLM step.

Run

From the repo root, after pip install -r python_skills/confirmed/x_pipeline/requirements.txt:

python python_skills/confirmed/x_pipeline/x_link.py LINK1 LINK2 --out x_inbox
python python_skills/confirmed/x_pipeline/x_link.py --links-file links.txt

Options: --model small|medium (whisper), --no-transcribe, --no-download, --no-articles, --cookies-browser chrome (protected posts).

Mode

Check mode until the user says this skill is established: read the script's output and x_inbox/SUMMARY.md and report anything missing. Once the user says it is established: run the script and do not read its source.

Output

x_inbox/notes/<id>.md, images/, videos/, SUMMARY.md, failed_*.txt. Exit code 0 = all captured, 1 = a link failed, 2 = no valid links.

Report back

Per link: captured or missing (text, article, images, video, transcript). Transcripts are machine-generated (faster-whisper), so names and numbers can be wrong; say so. Never paste keys or cookies into chat or the repo.