kiselev-read

Choose The Workflow

1. For quick latest-message or live probe requests, use references/low-level-ops.md.
2. For explaining one message or one person's latest message, use references/explain-message.md.
3. For day, week, or month analytics and timelines, use references/analytics.md.
4. For links, GitHub, themes, people, or focused extraction, use references/extract-artifacts.md.
5. For message schema, daily snapshot layout, and combined dataset rules, use references/data-contracts.md.
6. For short real phrasing and command normalization, use references/examples.md.
7. For the fixed weekly DEKSDEN report layout triggered by kiselev-week, use references/kiselev-week.md.

Required Operating Rules

• If the user explicitly says kiselev-read or kiselev-week, never substitute ИИшница or ai_club.
• Treat kiselev-week as a weekly-report alias inside this skill.
• When the user says kiselev-week, use the saved weekly report template from references/kiselev-week.md.
• Inside this skill, оба источника, две ленты, вместе, or канал + чат always mean: @deksden_notes + DEKSDEN (chat).
• Use only the default telega session profile for live fetch work.
• Keep the skill read-only end to end.
• Prefer live fetch evidence before abstract organization when the user asks for latest state.
• Treat channel and chat as one analytical space by default, but keep source labels visible.
• Deduplicate mirrored channel posts inside the chat by default when building combined datasets.
• Use thread-first logic for explanation: if a message is a reply, reconstruct the local thread first; otherwise inspect nearby messages and only then fall back to standalone reading.
• Keep answers in simplified Russian unless the user asks otherwise.

Core Commands

Quick live fetch from the public channel:

cd /home/pets/zoo/cc_chanels_telegram
./TOOLS/telega fetch --profile default @deksden_notes --limit 5 --json

Organize day or week snapshots with the working default session:

cd /home/pets/TOOLS/aiclub_skill
python3 scripts/analyze/fetch_kiselev_read.py --scope combined --period day
python3 scripts/analyze/fetch_kiselev_read.py --scope combined --period week

Build a deduped combined dataset from fetched snapshots:

cd /home/pets/TOOLS/aiclub_skill
python3 scripts/reporting/build_combined_snapshot.py --from 2026-03-19 --to 2026-03-25

Example User Messages

Use phrasing like this:

kiselev-read: забери сегодня и чат и канал
kiselev-read: покажи последние сообщения из чата и канала
kiselev-read: сначала забери потом проанализируй оба источника за сегодня простым русским
kiselev-read: забери неделю по каналу и чату
kiselev-read: сделай недельный отчёт по чату и каналу вместе
kiselev-week: сделай недельный отчёт
kiselev-week: сначала забери неделю и собери отчёт
kiselev-read: хочу таймлайн за 1 неделю по двум источникам
kiselev-read: собери ссылки за неделю из чата и канала
kiselev-read: github за неделю по чату и каналу вместе

Notes

• Fetch evidence shows that channel posts are mirrored into the chat nearly 1:1, so raw combined analysis must dedupe those mirrored posts by default.
• Use the public channel as the source of truth for original posts.
• Use the chat as the source of discussion, replies, and follow-up signals.
• If the user asks for оба источника, use DEKSDEN sources by default: @deksden_notes and DEKSDEN (chat).