I told everyone this, right?

@third_leademption

3.5K followers · usually 540 views a post

Views

4.9K

Likes

233

Comments

SoundTikTok sound · original sound - third_leademption

Promoted product

Discord

Why it worked

The video addresses a common pain point for developers and teams working with AI agents: data security and efficient collaboration. By offering a practical, safer alternative to traditional methods like GitHub sharing, it provides immediate value and a clear solution.

Summary

The creator explains a new, safer method for sharing an AI agent's information with a team, moving away from direct GitHub repository sharing. They demonstrate using a separate Discord server for this purpose, highlighting its ability to handle briefs and sources to generate content.

Structure

1Problem with sharing agent info via GitHub
2Introducing a safer solution using Discord
3Demonstrating Discord channels for agent interaction
4Explaining the agent's function within Discord
5Showing the code and current state of the solution
6Concluding with the benefits of the new method

Product placement

Discord: It acts as a communication platform where the creator has set up separate channels for interacting with an AI agent. Removing it would change how the creator communicates with the agent. GitHub: It is mentioned as a platform for sharing code and agent information with the team. Removing it would change how the creator shares information. Claude AI: Mentioned in the caption as an AI tool. It is not shown acting or being used in the video, so it merely appears. Removing it would not change what happens in the video.

Original caption

อันนี้ยังเป็นเชิงไอเดีย ที่เรากำลังพัฒนาอยู่ แน่นอนว่ามันเพิ่งเริ่ม… . แต่เป็นอีกวิธีที่ป้องกันข้อมูลหลุด และยังสามารถพัฒนาร่วมกันได้… . ยังไงขอไปพัฒนาต่อก่อน มันยังไม่เฟี้ยวเลย #Leademption #การตลาดออนไลน์ #Agent #ClaudeAI #GitHub

More from @third_leademption

On-screen text

Lab Note #65 ใช้ Agent ยังไงให้ข้อมูลไม่หลุด I told everyone this, right? That I share my Agent With the team through GitHub I just found a new way That's way safer I mean, I do trust my team But for other people Who don't have a team they trust Or someone who just joined Sharing a GitHub repo It's like handing over your whole brain They can just copy the files And you'd never trace who did it One Cmd+C, Cmd+V And it's gone They get the whole company's data A repo you share like that So I just found a new solution This is Discord I normally use it to talk to my team But I made a separate Discord Just to talk with the Agent It takes in briefs and sources And turns them into all kinds of content Making content, products, campaigns Turning ads on and off Making and tweaking creatives And posting too Welcome to #abby-briefing channel! This is the start of the #abby-briefing channel. Edit Channel Lab Note #65 7 มิถุนายน 2569 Avy online - Avy สรุปวันนี้ (07/06/2026 15:58) ด้าน Fix database connection สำหรับ Tasks DB (database ID ไม่ match) Fix query error ใน Content Tracker - ปัญหา Databases แนะนำประจำวัน อันนี้เป็นการคุยกันกับ drafts อีกครั้งนึง And it's all just a conversation Avy online - Avy สรุปวันนี้ (07/06/2026 16:03) ด้าน Fix database connection สำหรับ Tasks DB (ID: 27fd6d7a-000b85a09cac) - ตรวจสิทธิ์การแชร์กับ "content-inbox" Fix query error ใน Content Tracker - ปัญหา Databases แนะนำประจำวัน เมื่อเกิดการตรวจสอบ tasks/drafts ไปหลังบ้าน connection is There's a lot going on in here. I'll show you an update later Lab Note #65 7 มิถุนายน 2569 Avy online - Avy สรุปวันนี้ (07/06/2026 15:58) ด้าน Fix database connection สำหรับ Tasks DB (database ID ไม่ match) Fix query error ใน Content Tracker - ปัญหา Databases แนะนำประจำวัน อันนี้เป็นการคุยกันกับ drafts อีกครั้งนึง And it's all just a conversation Avy online - Avy สรุปวันนี้ (07/06/2026 16:03) ด้าน Fix database connection สำหรับ Tasks DB (ID: 27fd6d7a-000b85a09cac) - ตรวจสิทธิ์การแชร์กับ "content-inbox" Fix query error ใน Content Tracker - ปัญหา Databases แนะนำประจำวัน เมื่อเกิดการตรวจสอบ tasks/drafts ไปหลังบ้าน connection is There's a lot going on in here. I'll show you an update later Claude Code #1 Lab Note #65 ใช้ Agent ยังไงให้ข้อมูลไม่หลุด Path strings: `(tool_use_json prompt (conversation.py))` - `shared/config.py` comment line 5 - `shared/memory.py` config line 5 - `abby-intake/entry/discord_bot.py` line 12, 124: `(any-intent)` - `entry scripts (smoke_test, draft_smoke_test, smoke_test_probe): path strings` config/docs (non-breaking) - `abby-intake/profile.toml, env.example, MAC_SETUP.md, README.md` - `production/profile.toml, production/README.md` - `ads-ops/env.example, ads-ops/profile.toml` - `deployments/README.md` - `launch service reference to MAC_SETUP.md: com.leademption.avy-intake` Phase 1: `tool_use_json prompt (conversation.py)` `Haiku wrap JSON to code block -> json.loads -> raw JSON to Discord` `l module-level constant raw_SYSTEM` `REPLY_TOOL = { "name": "reply", "description": "rounds third-party memory (short)", "type": "object", "properties": { "type": {"type": "string", "description": "Discord message type"}, "text": {"type": "string", "description": "Discord message text"} }, "required": ["text"] } Right now it's still pretty rough `raw_SYSTEM` `use_block (DB times 41-44, 64-65): raw_tool_format: `(tool_use_json prompt (conversation.py))` line 67 `(memory field instruction)`: `(tool_use_json prompt (conversation.py))` line 1: `(tool_use_json prompt (conversation.py))` `use chat()`: `(messages create + parse)` `resp = client.messages.create( model=config.MODEL_ROUTINE, system=system_system, messages=messages ) tools = REPLY_TOOL tool_choice = "reply" (h for b in resp.content if get_attr(b, "type") == "tool_use")` And this is the solution I built To make it a good experience Your data stays safe And you can still direct it well Let's see how far I can take it

Transcript

I told everyone this, right? That I share my Agent With the team through GitHub I just found a new way That's way safer I mean, I do trust my team But for other people Who don't have a team they trust Or someone who just joined Sharing a GitHub repo It's like handing over your whole brain They can just copy the files And you'd never trace who did it One Cmd+C, Cmd+V And it's gone They get the whole company's data A repo you share like that So I just found a new solution This is Discord I normally use it to talk to my team But I made a separate Discord Just to talk with the Agent It takes in briefs and sources And turns them into all kinds of content Making content, products, campaigns Turning ads on and off Making and tweaking creatives And posting too Welcome to #abby-briefing channel! This is the start of the #abby-briefing channel. Lab Note #65 7 มิถุนายน 2569 Avy online - Avy สรุปวันนี้ (07/06/2026 15:58) ด้าน Fix database connection สำหรับ Tasks DB (database ID ไม่ match) Fix query error ใน Content Tracker - ปัญหา Databases แนะนำประจำวัน อันนี้เป็นการคุยกันกับ drafts อีกครั้งนึง And it's all just a conversation Avy online - Avy สรุปวันนี้ (07/06/2026 16:03) ด้าน Fix database connection สำหรับ Tasks DB (ID: 27fd6d7a-000b85a09cac) - ตรวจสิทธิ์การแชร์กับ "content-inbox" Fix query error ใน Content Tracker - ปัญหา Databases แนะนำประจำวัน เมื่อเกิดการตรวจสอบ tasks/drafts ไปหลังบ้าน connection is There's a lot going on in here. I'll show you an update later Claude Code #1 Lab Note #65 ใช้ Agent ยังไงให้ข้อมูลไม่หลุด Path strings: `(tool_use_json prompt (conversation.py))` - `shared/config.py` comment line 5 - `shared/memory.py` config line 5 - `abby-intake/entry/discord_bot.py` line 12, 124: `(any-intent)` - `entry scripts (smoke_test, draft_smoke_test, smoke_test_probe): path strings` config/docs (non-breaking) - `abby-intake/profile.toml, env.example, MAC_SETUP.md, README.md` - `production/profile.toml, production/README.md` - `ads-ops/env.example, ads-ops/profile.toml` - `deployments/README.md` - `launch service reference to MAC_SETUP.md: com.leademption.avy-intake` Phase 1: `tool_use_json prompt (conversation.py)` `Haiku wrap JSON to code block -> json.loads -> raw JSON to Discord` `l module-level constant raw_SYSTEM` `REPLY_TOOL = { "name": "reply", "description": "rounds third-party memory (short)", "type": "object", "properties": { "type": {"type": "string", "description": "Discord message type"}, "text": {"type": "string", "description": "Discord message text"} }, "required": ["text"] } Right now it's still pretty rough `raw_SYSTEM` `use_block (DB times 41-44, 64-65): raw_tool_format: `(tool_use_json prompt (conversation.py))` line 67 `(memory field instruction)`: `(tool_use_json prompt (conversation.py))` line 1: `(tool_use_json prompt (conversation.py))` `use chat()`: `(messages create + parse)` `resp = client.messages.create( model=config.MODEL_ROUTINE, system=system_system, messages=messages ) tools = REPLY_TOOL tool_choice = "reply" (h for b in resp.content if get_attr(b, "type") == "tool_use")` And this is the solution I built To make it a good experience Your data stays safe And you can still direct it well Let's see how far I can take it