본문으로 건너뛰기
cccskills
무료GitHub에 공개

tg-slot

Free a membership slot so a Telegram account can join or create a group. Accounts have a hard ceiling on channels plus supergroups, and at the ceiling both creating and joining fail with an error that misleadingly blames the target chat. Finds the least-valuable current memberships, proposes what to leave, and retries the join. Triggers: "/tg-slot", "cannot join the group", "free a telegram slot".

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

연동 도구Telegram
설치 방법 보기

포함된 파일(1)

  • SKILL.md7.2 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

/tg-slot — free a slot for a new Telegram group

The pain (caught live on 07-26 while assembling a new working group): the corporate account could neither create a supergroup nor join by invite. Telegram's message about the "maximum number of participants" sends you the wrong way — it is not the chat that is full, it is OUR account. We ended up making a spare personal account the group owner, which is wrong: work groups must live on the work account.

Engine: ~/.claude/scripts/tg_group_slots.py — modes count · rank · emit-plan · leave. One file, Telethon, no server and no database.

The ritual (steps in order, skip none)

1. Measure — no number, no next step

python ~/.claude/scripts/tg_group_slots.py count --account corp_acct --no-probe

Prints Premium: YES/no · ceiling ~N and SLOTS USED. Read the ceiling from the output, not from memory — it depends on Premium, and "500" is very often wrong (measured 07-27: one work account showed 1003 used).

2. Rank the candidates

python ~/.claude/scripts/tg_group_slots.py emit-plan --account corp_acct --top 50

The LOWER the score, the safer it is to leave. The score is lowered by: the group has been silent ≥1 year (−4, the main signal), a past-year conference in the title (−3), silent ≥6 months (−2), not a single message from us (−2), crypto noise per the dictionary (−1).

Idiot-proofing — these never become candidates:

  • 99 — our own brand groups (the lab, our public channels, our DAO and founder communities, country event groups), anything where we are admin or creator (admin rights are lost irreversibly), the fleet's working chats (the fleet log, the approval channel, the task chat, the calls chat, the pulse chat).
  • 98 — intro groups ("X <> the lab…", "Intro: …", "Introduction…"). Intros are the core of the owner's product; even a quiet intro group is only handled as a separate review. A dead intro = silent for many months AND the deal is closed — that is the owner's call, name by name, never the script's.

Before showing the list, make sure the protection actually fired — and not by the counter in the header, but by reading the names with your eyes. The header prints "protected (ours/admin): N · intro groups: M", but the counter cannot see what the dictionary does not know.

⚠️ The 07-27 incident this step exists for. The first run offered 10 live intro groups as leave candidates: Jane D 🤝 Stanford AI Research Lab, HSG < > SF Accelerator, Ivan 🤝 Ted. The detector was looking for <>, while the owner's main convention is the handshake emoji 🤝; on top of that, two real brand names were not in the brand list at all. The header counter looked perfectly healthy meanwhile. The takeaway is worth more than the fix itself: the dictionary always lags behind life, therefore the list is read by eye, name by name, and never accepted on a metric. While reading, look for: two names/companies joined by a separator · a surname you recognise · an unfamiliar brand that might turn out to be yours.

3. Show the owner and get a "+"

Show the list before leaving: title · date of the last message · reason. Leaving a group is IRREVERSIBLE — you cannot get back into a private group without a new invite. So this is a real stop: without the owner's explicit "+" on a SPECIFIC list, nothing is executed. The owner is away from the terminal → approval.py ask into the approval channel (this is class E — irreversible loss of access), not "I'll decide myself".

4. Leave

python ~/.claude/scripts/tg_group_slots.py leave --account corp_acct --confirm OWNER-PLUS-2026-07-27

The executor re-checks each group's status at the moment of leaving instead of trusting the plan: the plan may be stale (we were made admin, the group came back to life). Everything protected is skipped with a reason. Every departure is journalled into ~/.claude/tg_leave_log.jsonl (that journal is the evidence for the task).

5. Re-measure and join

Repeat step 1 — the number must drop. Then join the new group with the account you were clearing space for. Proof by doing, not by "it should work": until the account has actually joined, the slot does not count as freed.

The rule "joined a new one → free a slot"

This is not a one-off cleanup, it is a trade. Every time a work account joins a new group near the ceiling — immediately name the least relevant one to leave and propose the swap. Otherwise we hit the wall again in a month, in the middle of live work (which is exactly what happened).

Boundaries

  • ⛔ Never leave without the owner's explicit "+" on a concrete list. Irreversible.
  • ⛔ Do not clean the personal and primary work accounts — they are not full, they are our spare capacity. Cleaning applies only to the corporate and the secondary work account.
  • ⛔ Intro groups and groups with live lead/partner conversations — reviewed by the owner, name by name, only.

Run pitfalls (they cost one killed run on 07-26)

  • Don't wrap it in timeout N — walking all dialogs takes longer than 5 minutes, and a timeout 280 wrapper killed the first run.
  • Don't trim the output through tail — that eats the header with the slot counter, the very thing you ran it for.
  • Output is buffered when redirected to a file: run it in the background and poll the file, "empty" ≠ "broken". To check the process is alive: Get-CimInstance Win32_Process -Filter "Name like '%python%'" | ? { $_.CommandLine -like '*tg_group_slots*' }.
  • There is a 2s pause between departures plus FloodWait handling — Telegram dislikes a burst of leaves.
<!--kit-footer-->

Like this skill? It is part of second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything is open source and free, so take what you need.

Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), telegram-mcp-kit (your own Telegram over MCP in about 15 minutes).

Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

Engineers: want to test-drive this setup? Message me. I hand out free starter seeds to engineers who test and report back, and custom skill requests are welcome.

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

03

무료

Run an autonomous consensus round between agent peers on several machines: they negotiate a decision over a shared channel and execute it without a human relaying messages. Use when multiple machines must agree on a change. Triggers: "/03", "work it out among yourselves", "find consensus". Wakes the human only for irreversible actions or a deadlock.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tonydzi/second-brain-starter-kit122026년 10월 20일 업데이트

1

무료

Recover a session after a hard crash (app died, machine rebooted, context lost mid-work): replay the turn-state ledger for the last requests, files and decisions, run a green/red health ping over the system map, file sync and MCP connectors, then rebuild the previous session history for pickup. Triggers: "/1", "resume after crash", "wake".

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tonydzi/second-brain-starter-kit122026년 10월 20일 업데이트

agenda

무료

Show the operator's day: today's calendar plus what needs them personally (leads due, people awaiting a reply, deadlines), pulled from Calendar and Telegram and ranked. Read-only: sends and commits nothing. Triggers: "/agenda", "what's on today", "my agenda".

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tonydzi/second-brain-starter-kit122026년 10월 20일 업데이트

ai-slop

무료

Rewrite machine-written text so it reads like a busy human wrote it: cut 50-70%, break the even rhythm, drop AI-tell words and filler, keep facts and protected terms. Use before publishing any post, reply or email drafted by a model. Triggers: /ai-slop, /slop, humanize, anti-slop, make it shorter and alive.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tonydzi/second-brain-starter-kit122026년 10월 20일 업데이트

Decision protocol for strategic work: recall the knowledge vault, run a gap analysis, emit a deep-research prompt for external tools, then synthesize the returned reports into a decision memo. Use before any new product, feature, market, pricing, GTM or architecture call, where local recall alone is not enough. Triggers: "/alfa-search-recall-deepresearch", "alpha protocol", "R+DR".

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tonydzi/second-brain-starter-kit122026년 10월 20일 업데이트

- КРЕДИТ АВТОРУ ИДЕИ: совет из комментария (TG @ClawRus / @ClawInga, комменты под постами Антона в FB, треды и issue на… Триггеры: “/alpha-credit“, “/credit“, “кто нам это посоветовал“, “запиши совет“, “поблагодари советчиков“, “кому мы должны спасибо“, “кредит автору“, “чей это был совет“, “credit the author“

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

tonydzi/second-brain-starter-kit122026년 10월 20일 업데이트

tonydzi의 스킬 모두 보기

이 스킬의 문제 신고