Send messages between agent sessions through a machine-wide SQLite message bus.
npx skills add https://github.com/Arasz/ai-badger --skill send-message-arasz
Please help me install this Agent Skill. Skill: send-message Source: https://github.com/Arasz/ai-badger/tree/main/features/common/skills/send-message Command: npx skills add https://github.com/Arasz/ai-badger --skill send-message-arasz
--session-id, project broadcasts via --project-id, and machine broadcasts when neither target is given; if both are given, the session target wins. - Mandatory sender identity: session and project ids are derived from harness environment variables, pid ancestry, and the .ai-badger/project-id walk, or passed explicitly; unresolvable identity refuses the send with a non-zero exit. - Target validation: a --project-id target is checked against projects resolvable on this machine before anything is written, so undeliverable rows are never stored. - Use Case: An agent finishes investigating a bug and runs the send script with another session's id and the message "found it, see src/bus.py"; the receiving session's delivery hook injects the message into its context on its next turn. ## Quick Start Ask the agent to send a message to another session by running the send_message.py script with --content and the target --session-id.High-intent search queries and answers about installing and using this skill.
Run send_message.py with --content and the target's --session-id for a 1:1 send. The script prints "sent <row id>" on success, and the receiving session's delivery hook injects the message into its context on its next delivery event.
Pass --project-id with the project's minted id instead of --session-id. The target is validated against projects resolvable on the machine before anything is written; omitting both target flags sends a machine-wide broadcast.
Both sender sessionId and projectId are required. The session id derives from harness env vars (CLAUDE_CODE_SESSION_ID, PI_SESSION_ID, HERMES_SESSION_ID), pid ancestry, or a unique cwd match; the project id comes from AI_BADGER_PROJECT_ID or the nearest .ai-badger/project-id file. Pass --sender-session and --sender-project explicitly when neither can be derived.
The session target wins: the row is stored as a 1:1 send and the project half is dropped at write time. Dual-flag sends also skip project-target validation since no project target is ever stored.
Yes, pi has a native message-bus extension tool that speaks to the same user-DB bus with automatic sender identity derivation. If the extension is not installed, fall back to the send_message.py script, which remains the transport for all other harnesses.
The machine scan validating targets is bounded: it walks at most four directory levels under the store's home, skips noise trees like node_modules and .git, and never follows symlinks. A project outside that budget is invisible; use the minted id from its .ai-badger/project-id file.
AraszSend messages between agent sessions through a machine-wide SQLite message bus.
AraszCoordinates parallel agent sessions over a project message bus with broadcast events and ack discipline.
Khang5687Track and report delegated task distribution, account usage, and model performance.
ruskibeatsDiagnose Pi extension loading failures by checking file structure, package.json, dependencies, and Pi configuration.