jetson-discord-scan

name	jetson-discord-scan
type	command
description	Read-only shallow scan of the Jetson AI Lab Discord — list PUBLIC channels, read a channel's recent messages, or find which public channels are active/high-traffic in the last month. Use when the user says "scan the Jetson AI Lab Discord", "what's active on Discord", "which channels are busy", "read

Jetson AI Lab Discord scan

Read-only reconnaissance of the Jetson AI Lab Research Group Discord (guild 1326246312072581160). This skill is a thin shim around the jlab CLI — the real implementation lives in the jlab discord subcommands. It never posts, reacts, or creates threads. This is the read/fetch side of the agent's intended job; the indexing and question-answering parts are not built yet — this skill only reads.

Public channels only. The bot token is read-scoped, and this skill filters to channels the guild's @everyone role can view. Private / role-gated channels are excluded from every default code path, so their names and contents never leak into a scan result or the committed channel map. (--all is the one opt-in that includes private channels, for an admin verifying the split.)

Prerequisites

• DISCORD_BOT_TOKEN exported in the environment (read-scoped bot token — read from the env, never passed as a flag).
• jlab CLI installed and on $PATH.

Run doctor first to confirm the setup:

jlab discord doctor
# -> {"ok": true, "guild_id": "1326246312072581160", ...}

Usage

# Which PUBLIC channels were active in the last month, ranked by traffic?
jlab discord active                                  # since=30d, all active channels
jlab discord active --since 7 --top 15 --preview 3   # last week, top 15, 3 msgs each
jlab discord active --since 30 --limit 100 --par 12  # deeper count, more parallelism

# List public channels (each tagged "public": true); --all adds private ones.
jlab discord channels
jlab discord channels --all

# Read one channel's recent messages (1-100, default 20).
jlab discord read 1427964860573945856 --limit 50     # e.g. #nvda-dgx-spark

All results are JSON on stdout; diagnostics/errors go to stderr. Exit codes follow the agent-first contract: 0 ok, 1 user-input error, 2 environment/setup error (missing token, absent CLI, unreadable guild).

active output shape

{
  "guild_id": "1326246312072581160",
  "since_days": 30,
  "fetch_limit": 40,
  "probed_text_channels": 97,
  "active_channels": 46,
  "channels": [
    {
      "id": "...", "name": "nvda-dgx-spark",
      "last_post": "2026-06-17T05:41:12+00:00",
      "msgs_in_window": 40,
      "saturated": true,          // >= fetch_limit posts in window — traffic underestimated; raise --limit
      "preview": [{"author": "...", "content":   
    
  

Channels are ranked by msgs_in_window (in-window traffic), then by last_post. saturated: true means the channel hit the fetch limit — bump --limit (max 100) for a truer count on the busiest channels.

How it works

This skill is a thin shim: it delegates to jlab discord <subcommand>. The jlab CLI handles all the heavy lifting — talking to the Discord API, filtering public channels, fanning out parallel reads, and merging results.

• jlab discord active probes each public text channel for recent messages, ranks them by traffic, and returns a ranked list with optional previews.
• jlab discord channels lists channels, tagging each with a public flag (a channel is public iff @everyone has view_channel).
• jlab discord read fetches recent messages from a single channel.

The jlab CLI reuses the discord-bot-cli library under the hood for the Discord transport layer, but agents should interact only through jlab discord subcommands.