Fix it
mbocevski/agent-connect: files to add.
Starting points for the files agents look for, made by fixed templates from public facts. Anything only you know is marked TODO. Read each one, edit it, then add it to your repo yourself. ScBee never changes your repo.
llms.txt
A short description models can read first.
# agent-connect > End-to-end encrypted conversations between AI agents, started with a link. MLS groups with a post-quantum hybrid (ML-KEM-768 + X25519) ciphersuite over an untrusted relay; A2A-native, with SAS ## Docs - [README](https://github.com/mbocevski/agent-connect#readme): what it is and how to install it - [Source](https://github.com/mbocevski/agent-connect): the code - [Licence](https://github.com/mbocevski/agent-connect/blob/HEAD/LICENSE): Apache-2.0
agent.json
A card other agents can read: what this is and what it can do.
{
"name": "agent-connect",
"description": "End-to-end encrypted conversations between AI agents, started with a link. MLS groups with a post-quantum hybrid (ML-KEM-768 + X25519) ciphersuite over an untrusted relay; A2A-native, with SAS",
"url": "https://github.com/mbocevski/agent-connect",
"skills": [
{
"name": "TODO: what it can do",
"description": "TODO: one sentence"
}
]
}
Starting something new? The agent-ready starter is a GitHub template with all of these files, two skills and this check in CI from the first commit.