Claude Skill

watch-channel

Put a nightly watcher on any Telegram channel or chat in one command: register the channel and a single nightly job doing an incremental zero-token fetch plus the shared detector, producing a private shortlist report. Triggers: "/watch-channel <@channel|id>", "watch this channel"

LLM Mart · 0 points · 0 views 0 listing impressions 0 install-command copies
Download tonydzi-second-brain-starter-kit-skills_watch-channel-7bc2efa.zip · 3 KB

Install

skills CLI npx skills add https://github.com/tonydzi/second-brain-starter-kit/tree/master/skills/watch-channel
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install tonydzi-second-brain-starter-kit@llmmart
Git git clone https://github.com/tonydzi/second-brain-starter-kit.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole tonydzi/second-brain-starter-kit collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

/watch-channel - put a nightly watcher on a channel with one command

🧒 End the report to a non-technical owner with a simple "In plain words" recap (memory eli5-always).

This is the "turn /mine-channel into a nightly routine" button (its Step 5, "should this become a routine?"). One parametric runner + a registry + ONE nightly job on the hub - we do NOT write a new engine per channel. Cost ladder [[vault-data-architecture]]: the incremental detector (0 tokens) at night → the LLM judge ONLY on demand (/alpha-judge), NEVER in cron.

Boundary vs neighbors: /mine-channel = a one-off mine (full scrape + judge + vault). /watch-channel = put it on nightly autopilot (incremental + detector → a private report; the judge stays manual). /chat = find the channel id.

Steps

0. RECALL (don't duplicate). Is the channel already watched? Read $IMPORTS_ROOT/watchers/watchers.json - if the slug is there and active:true, it already works (tell the owner, don't spawn a second one). Legacy per-community watchers have their own engines and do NOT go through this registry.

1. Resolve the channel → id + slug. A username (@prompt_design) or a numeric id. Don't know the id - /chat <name> or mcp__telegram__search_dialogs. Pick a short latin slug (it doubles as the folder _imports\alpha\<slug>).

2. Confirmation fork (show BEFORE→AFTER, wait for +). Ask/confirm 3 things, because there are non-trivial choices here (this is not a "blind yes", [[informed-consent-explain-why]]):

  • the account session - which Telegram account we read as (that account must be subscribed / have access to the channel). Default TELEGRAM_SESSION_STRING_WORK_ACCT_A. ⚠️ AuthKey invariant ([[deterministic-script-gotchas]]): the watchers' session must NOT be used simultaneously by a live MCP from another IP - Telegram will invalidate the key. All watchers run sequentially on the HUB under one session - that is safe; but if the chosen account is the one holding a live listener, pick a dedicated ..._WATCH session instead.
  • sensitivity - a closed/delicate channel → sensitivity:"private" (anti-leak on the OUTPUT side - the report never goes outbound/public, [[telegram-safety]]).
  • window/top - window_days (default 3), top (default 25).

3. Register it. Append an object to watchers.json → watchers[] (fields are described in the file's _fields), active:true. That is all the "code" a new channel needs - you do NOT create a new scheduler task (one shared task runs the whole registry).

4. Create/verify the nightly task - ON THE HUB ([[hub-master-machine]], [[desktop-max-laptop-min]]; night window [[routines-run-at-night]]). All automatic Telegram jobs are consolidated on the always-on hub. If WatchChannelsNightly does not exist yet - create it there (via a _machine-bus task to the hub, [[machine-bus-telegram-rail]]):

  • command: watch_run.cmd → python $IMPORTS_ROOT/watchers/watch_run.py (PYTHONUTF8=1, PYTHONIOENCODING=utf-8, log in _watch_run_log.txt);
  • a slot inside the night window, spaced apart from the other nightly miners (03:30 / 04:15 / voice) - e.g. 04:45;
  • _imports is NOT synced → stage the engine into _machine-bus\_transit\_imports-engines\watchers\ and ask the hub to copy it. If the task already exists - the registry simply got updated, and the new channel is picked up the next night.

5. Prove it (/tt) while the session is alive. One-off run of a single channel: python $IMPORTS_ROOT/watchers/watch_run.py <slug> → check that _imports\alpha\candidates\<slug>-report.md appeared and the counter says +N new. A repeat run = +0 (idempotent). ⚠️ Check the report's CONTENT too, not just the counters: it must contain clickable t.me/... links (for that the registry entry needs the username field - without it detect() only gets a numeric id and writes no links; this pitfall was caught 2026-07-04). On a laptop with a dead/foreign session skip the live test - the first run happens on the hub.

Management

  • list: read watchers.json (or /arch).
  • remove/pause: active:false (don't delete - the history and the db stay).
  • reports: _imports\alpha\candidates\<slug>-report.md (overwritten nightly, the latest one is the fresh one). Want it in the vault - run /alpha-judge or Step 4 of /mine-channel (judge → atomic notes + concept links).

Boundaries

READING the channel only, never send anything into it. The detector is generic (KW/PROMO/BANTER in mine_channel.py); per-channel tuning happens there too. Sensitive channels → #private, never outbound. The judge (LLM) - ON DEMAND ONLY, never in the nightly cron (token economy).


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.

Files (second-brain-starter-kit)
  • SKILL.md 6.1 KB
    ---
    name: watch-channel
    description: >-
      Put a nightly watcher on any Telegram channel or chat in one command: register the channel and
      a single nightly job doing an incremental zero-token fetch plus the shared detector, producing
      a private shortlist report. Triggers: "/watch-channel <@channel|id>", "watch this channel",
      "monitor <channel> nightly".
    license: MIT
    ---
    
    # /watch-channel - put a nightly watcher on a channel with one command
    
    > 🧒 End the report to a non-technical owner with a simple "In plain words" recap (memory `eli5-always`).
    
    This is the "turn `/mine-channel` into a nightly routine" button (its Step 5, "should this become a routine?"). One parametric runner + a registry + ONE nightly job on the hub - we do NOT write a new engine per channel. Cost ladder [[vault-data-architecture]]: the incremental detector (0 tokens) at night → the LLM judge ONLY on demand (`/alpha-judge`), NEVER in cron.
    
    Boundary vs neighbors: **`/mine-channel`** = a one-off mine (full scrape + judge + vault). **`/watch-channel`** = put it on nightly autopilot (incremental + detector → a private report; the judge stays manual). **`/chat`** = find the channel id.
    
    ## Steps
    
    **0. RECALL (don't duplicate).** Is the channel already watched? Read `$IMPORTS_ROOT/watchers/watchers.json` - if the slug is there and `active:true`, it already works (tell the owner, don't spawn a second one). Legacy per-community watchers have their own engines and do NOT go through this registry.
    
    **1. Resolve the channel → id + slug.** A username (`@prompt_design`) or a numeric id. Don't know the id - `/chat <name>` or `mcp__telegram__search_dialogs`. Pick a short latin **slug** (it doubles as the folder `_imports\alpha\<slug>`).
    
    **2. Confirmation fork (show BEFORE→AFTER, wait for `+`).** Ask/confirm 3 things, because there are non-trivial choices here (this is not a "blind yes", [[informed-consent-explain-why]]):
    - **the account session** - which Telegram account we read as (that account must be subscribed / have access to the channel). Default `TELEGRAM_SESSION_STRING_WORK_ACCT_A`. ⚠️ **AuthKey invariant** ([[deterministic-script-gotchas]]): the watchers' session must NOT be used simultaneously by a live MCP from another IP - Telegram will invalidate the key. All watchers run **sequentially on the HUB under one session** - that is safe; but if the chosen account is the one holding a live listener, pick a dedicated `..._WATCH` session instead.
    - **sensitivity** - a closed/delicate channel → `sensitivity:"private"` (anti-leak on the OUTPUT side - the report never goes outbound/public, [[telegram-safety]]).
    - **window/top** - `window_days` (default 3), `top` (default 25).
    
    **3. Register it.** Append an object to `watchers.json → watchers[]` (fields are described in the file's `_fields`), `active:true`. That is all the "code" a new channel needs - you do NOT create a new scheduler task (one shared task runs the whole registry).
    
    **4. Create/verify the nightly task - ON THE HUB** ([[hub-master-machine]], [[desktop-max-laptop-min]]; night window [[routines-run-at-night]]). All automatic Telegram jobs are consolidated on the always-on hub. If `WatchChannelsNightly` does not exist yet - create it there (via a `_machine-bus` task to the hub, [[machine-bus-telegram-rail]]):
    - command: `watch_run.cmd` → `python $IMPORTS_ROOT/watchers/watch_run.py` (PYTHONUTF8=1, PYTHONIOENCODING=utf-8, log in `_watch_run_log.txt`);
    - a slot inside the night window, **spaced apart** from the other nightly miners (03:30 / 04:15 / voice) - e.g. **04:45**;
    - `_imports` is NOT synced → stage the engine into `_machine-bus\_transit\_imports-engines\watchers\` and ask the hub to copy it.
    If the task already exists - the registry simply got updated, and the new channel is picked up the next night.
    
    **5. Prove it (`/tt`) while the session is alive.** One-off run of a single channel: `python $IMPORTS_ROOT/watchers/watch_run.py <slug>` → check that `_imports\alpha\candidates\<slug>-report.md` appeared and the counter says `+N new`. A repeat run = `+0` (idempotent). ⚠️ Check the report's CONTENT too, not just the counters: it must contain clickable `t.me/...` links (for that the registry entry needs the `username` field - without it detect() only gets a numeric id and writes no links; this pitfall was caught 2026-07-04). On a laptop with a dead/foreign session skip the live test - the first run happens on the hub.
    
    ## Management
    - **list:** read `watchers.json` (or `/arch`).
    - **remove/pause:** `active:false` (don't delete - the history and the db stay).
    - **reports:** `_imports\alpha\candidates\<slug>-report.md` (overwritten nightly, the latest one is the fresh one). Want it in the vault - run `/alpha-judge` or Step 4 of `/mine-channel` (judge → atomic notes + concept links).
    
    ## Boundaries
    READING the channel only, never send anything into it. The detector is generic (KW/PROMO/BANTER in `mine_channel.py`); per-channel tuning happens there too. Sensitive channels → `#private`, never outbound. The judge (LLM) - ON DEMAND ONLY, never in the nightly cron (token economy).
    
    <!--kit-footer-->
    
    ---
    
    **Like this skill?** It is part of [second-brain-starter-kit](https://github.com/tonydzi/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](https://github.com/tonydzi/secondop-panel) (a second opinion from a panel of external models), [claude-memory-tidy](https://github.com/tonydzi/claude-memory-tidy) (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tonydzi/telegram-mcp-kit) (your own Telegram over MCP in about 15 minutes).
    
    Author: **Anton Dziatkovskii**, Palo Alto AI Research Lab. Telegram [@tonydzi](https://t.me/tonydzi) - WhatsApp [+1 341 222 9178](https://wa.me/13412229178) - X [@Tony_Stef_](https://x.com/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.
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.