agentleFS
matthewdonsemail-labmatthewdonsemail-lab/log/.claude/skills/log-discord-group-chat-read/SKILL.md

log-discord-group-chat-read

Skill63 starsChanged 2 days agoMITView on GitHub

Reading a Discord group chat without Twenty. A log play (docs/useCases/discord/groupChatRead.md): Catch up on what was said in a Discord group DM over the last few hours, on the account's own session, with the account kept in a local file instead of Twenty. This page reads and sends nothing. A channel inside a server is read the same way by `\"<server> #<channel>\"`, but always live from Discord, since log.db keeps only its mentions and replies: see [serverReply](serverReply.md) for how a server channel differs from a group DM.

The file

Files
SKILL.md

Reading a Discord group chat without Twenty

This is the log play discord/groupChatRead. Before anything else, call the log MCP tool getGuide({ useCase: "discord/groupChatRead" }): it returns this page and makes it the session's play, so every log tool's answer says which step you are on and a send with no play loaded is refused (rules 157, 158). Then follow the steps in order.

Draft every message from that person's own thread and matt's real replies (voice with no text, search), never from memory, and never send two that read alike (sameAsSent).

Catch up on what was said in a Discord group DM over the last few hours, on the account's own session, with the account kept in a local file instead of Twenty. This page reads and sends nothing. A channel inside a server is read the same way by "<server> #<channel>", but always live from Discord, since log.db keeps only its mentions and replies: see serverReply for how a server channel differs from a group DM.

How the owner asks for it: "read the <group> chat over the last <N> hours", "what did <someone> set up in <group>"

Run it from a fresh context

Say: "go on docs/useCases/discord/groupChatRead.md for <group>, last <N> hours". The agent reads this page and runs the steps.

The steps

#StepToolDone when
0The account is known: a local row if it isn't in TwentylocalAccountAdd { platform: "discord", account: "<account>", adsPower: "<profileId>" }log localAccount list shows it
1Session checkdiscordSessionCheckloggedIn
2If logged out: open the account's own browser, the owner signs in, then stop it to save the sessioninspectStart { account }, then inspectDoLive stopthe session check answers loggedIn
3Find the group or channel by its namediscordChatsthe group by its name, with its members; a server channel as "<server> #<channel>"
4Read back far enough for the window, the chat named as the owner says it, never its iddiscordRead { account, channel: "<name>", limit }the oldest message read is older than the window, or the answer says stopped
5Summarise by who said what and what was decided, quoting timesthe reply to the ownerevery claim names its message time and author

A busy group runs about 100 messages an hour, so a 5 hour window needs a limit near 500 to 800 (the cap is 1000). --after filters within the messages read, it doesn't reach further back.

What the runs taught

  • Twenty doesn't have to hold the account. LOG_ACCOUNTS=local leaves Twenty out entirely, and a session captured by inspectStart saves back to the local file.
  • A 429 stops the read. The answer keeps what was read and says stopped, with Discord's status, and a 429 pauses the account. Nothing retries (rule 94), and no script calls Discord with the account's token. Any refusal stops log search index across every chat the same way, keeping what it had read.
  • A browser profile that won't start for low disk needs space freed and the browser app restarted, never its cache deleted: that can wipe the session.
  • A chat is named, never passed by id. A discordRead with a bare id is refused by the chat-names hook: pass "A", "aeo" or "<server> #<channel>", and the tool finds the id in the kept chat list.
  • The read comes from the .db when it is current. Once the account has caught up and the chat was backfilled, discordRead answers from the .db in milliseconds and says so; Discord is read only to fill a chat the .db doesn't hold yet (rule 160).
  • A message with no text is never answered blind. An empty message with no attachment saved (a sticker, a call, an embed) is told to the owner as it is, and nothing is drafted until it can be placed.
  • When the owner is live in the chat, leave it to him. If his own message comes after theirs, he is already answering: release the claim and send nothing. A question only he can answer (when he's flying home, a date, a plan he hasn't said) waits for him, whoever is reading.
  • Secrets posted in the chat (keys, tokens) are never repeated in the summary; tell the owner to rotate them.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.