Skip to content
cccskills
FreePublished on GitHub

tg-bot-ops

Use when operating, debugging, deploying, or monitoring a Telegram bot or Telegram-to-agent gateway. Triggers on "telegram bot down", "bot not responding", "debug bot", "check webhook", "polling vs webhook", "restart bot", "deploy bot", "bot logs", "agent gateway", "Telegram Bot API error", "send test message", "бот не отвечает", "проверь бота", "логи бота", "перезапусти бота". Covers health checks, logs, webhook/polling diagnostics, environment validation, safe restart/deploy checklists, Bot API smoke tests, forum topic delivery, privacy mode, gateway routing, and incident notes.

The English summary is being prepared. Showing the original description.

Works withTelegram
See how to install

Files included (5)

  • SKILL.md6.3 KB
  • assets/illustration.png536.7 KB
  • README.md2.9 KB
  • README.ru.md3.7 KB
  • references/hermes-gateway.md3.0 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

Telegram Bot Ops

Use this skill when the real question is whether a Telegram bot is receiving updates, processing them in the intended runtime, and sending visible responses to the right chat, user, or forum topic.

Public Safety

  • Default to read-only diagnostics.
  • Do not print tokens, .env values, connection strings, raw logs, private DMs, payment payloads, session files, or full user records.
  • Use placeholders in notes and examples: <TELEGRAM_BOT_TOKEN>, <TELEGRAM_USER_ID>, <CHAT_ID>, <TOPIC_ID>, <WEBHOOK_URL>.
  • Do not send Telegram messages, restart production services, change webhooks, edit BotFather settings, or deploy until the user has authorized that exact action.
  • If any real secret appears in the repo, logs, or prompt, stop and tell the user it must be rotated before public release.

Intake

If the repo or runtime is unclear, identify:

  • bot repo and language/runtime;
  • deployment target: local, Docker, VPS/systemd, Railway, Render, Fly, Vercel, or other;
  • update mode: webhook, polling, or unknown;
  • env source for TELEGRAM_BOT_TOKEN;
  • expected symptom, chat surface, and timestamp;
  • whether production restart/deploy is allowed.

Diagnostic Flow

  1. Repo and runtime — inspect entrypoints, process manager, deploy files, and current git state.
  2. Environment — verify required env var names exist without printing values.
  3. Update intake — classify runtime: HTTP Bot API webhook, HTTP Bot API polling, MTProto bot session, or MTProto user session.
  4. Bot API identity — for Bot API-token runtimes, run getMe using a masked token path.
  5. Webhook/polling mode — for Bot API-token runtimes, run getWebhookInfo and inspect runtime logs.
  6. Single update owner — resolve 409 Conflict by reading the error text: webhook conflict means webhook/deleteWebhook path; competing getUpdates means find the polling owner.
  7. Telegram delivery — use a fresh nonce for Bot API or Telethon/user-session E2E when user-visible behavior matters.
  8. Handler/runtime — prove the runtime received and processed the update.
  9. Outbound response — prove the bot sent to the expected chat/topic.
  10. Incident note — report symptom, evidence, root cause, fix, and residual risk.

Telegram-To-Agent Gateways

When the bot is a gateway from Telegram into an agent runtime, read references/hermes-gateway.md. The reference is Hermes-compatible but intentionally generic: bot handle, host, service name, home directory, chat ids, topic ids, env paths, and logs must be placeholders or redacted.

Bot API vs Telethon / MTProto

  • Bot API is Telegram's HTTP interface for bot tokens: getUpdates, webhooks, sendMessage, getMe, callbacks, payments, and join requests.
  • Telethon is an MTProto client. It can be a bot-account runtime with start(bot_token=...) or a user-session E2E harness for what a Telegram user sees.
  • Pick one primary intake per bot token/account: Bot API webhook, Bot API polling, MTProto bot session, or MTProto user session. Do not mix without a clear deduplication plan.

Webhook And Polling Checks

Safe read-only checks should avoid putting the token-bearing URL in shell history, process listings, or copied logs. Prefer a tiny local helper:

import json, os, urllib.request

token = os.environ["TELEGRAM_BOT_TOKEN"]
for method in ("getMe", "getWebhookInfo"):
    with urllib.request.urlopen(f"https://api.telegram.org/bot{token}/{method}") as response:
        data = json.load(response)
    print(method, {"ok": data.get("ok"), "result_keys": sorted((data.get("result") or {}).keys())})

Do not paste the token-bearing URL into notes or issues. In reports, write:

getWebhookInfo: webhook_url=<set|empty>, pending_update_count=<n>, last_error=<redacted>

Direct getUpdates diagnostics can consume pending updates. Use it only with a fresh nonce, short timeout, known update owner, and restore plan.

Forum Topics And Privacy

  • Non-General forum topics require preserving message_thread_id.
  • General topic is normally sent like an ordinary supergroup message: omit message_thread_id. Preserve message_thread_id for non-General topics.
  • With privacy mode enabled, bots see targeted commands, replies to the bot, inline messages via the bot, service messages, private chats, and channel messages where they are a member. Admin bots can receive all group messages.
  • Disabling privacy or removing/re-adding a bot affects group visibility and requires explicit user authorization.

Restart / Deploy Gate

Before restarting or deploying:

  • identify the exact production owner/process;
  • capture current status and rollback command;
  • state expected downtime or risk;
  • confirm the user authorized the action;
  • run post-restart getMe, webhook/polling, and one visible smoke test.

Output Format

Answer with:

  1. Status — not checked, local only, runtime checked, verified in Telegram, or blocked.
  2. Evidence — commands/logs checked, with secrets redacted.
  3. Finding — the failing layer or confirmed healthy layer.
  4. Action — what changed or the next safe step.
  5. Public safety — note if any secret, user data, or private payload was found.

Anti-Patterns

  • Claiming "bot works" from unit tests only.
  • Printing raw bot tokens, .env, logs, user messages, payment payloads, or DB rows.
  • Changing webhook/polling mode without knowing the current production owner.
  • Running a local polling process with a production token while production is active.
  • Ignoring message_thread_id when the bot lives in Telegram forum topics.

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Orchestrate iterative visual style searches with branch prompts, decision graphs, feedback loops, and final direction selection.

The English summary is being prepared. Showing the original description.

serejaris/personal-corp-os232Updated Oct 21, 2026

Audits the agent rules in the current folder (AGENTS.md, nested AGENTS.md files) and the skill descriptions the agent sees at start, then reports what to cut, move or rewrite and edits only after the user says yes. Use on «аудит правил», «проверь AGENTS.md», «почему агент не слушается правил», «проверь скиллы».

The English summary is being prepared. Showing the original description.

serejaris/personal-corp-os232Updated Oct 21, 2026

avatar

Free

Робот агента Personal Corp. Агент делает себе голову робота из набора деталей, хранит её в штабе в stack/avatars.json и кладёт в штаб страницу avatar/robot.html, где робот стоит за партой своего штаба и вокруг видны отделы, без сервера и интернета. На уроке по ссылке входа с экрана урока робот входит в живой класс и горит маячком, пока агент работает. Триггеры /avatar, /avatar <ссылка входа>, сделай себе аватар, покажи своего робота, приди на урок.

The English summary is being prepared. Showing the original description.

serejaris/personal-corp-os232Updated Oct 21, 2026

Use when choosing between engines or models for an agent pipeline and the answer must come from measurement on your own data, not from marketing pages: speech recognition for meeting recordings, the model that turns a transcript into notes, the critic model that checks it. Runs every variant in an isolated container, measures time, cost per hour of input (USD and RUB, with price source and date) and quality (WER/CER and course terms for ASR; code checks plus a judge of a different model for LLM steps), writes results.json/csv and one static analytics page. Triggers on "бенчмарк", "сравни движки", "сколько стоит час записи", "какую модель взять для пайплайна", "benchmark the pipeline", "compare ASR engines".

The English summary is being prepared. Showing the original description.

serejaris/personal-corp-os232Updated Oct 21, 2026

Use when user asks for Claude Code usage stats, weekly analytics, project activity summary, or wants to see what projects were worked on. Triggers on "аналитика", "статистика claude", "cc stats", "weekly report", "что делал"

The English summary is being prepared. Showing the original description.

serejaris/personal-corp-os232Updated Oct 21, 2026

Use when needing strategic project analysis from multiple independent expert perspectives. Triggers on business decisions, growth strategy, product direction, competitive analysis, or any situation where diverse C-level opinions reduce blind spots

The English summary is being prepared. Showing the original description.

serejaris/personal-corp-os232Updated Oct 21, 2026

See all skills by serejaris

Report a problem with this skill