from tinykaggleclaw48
Rules and workflows for messaging, delegation, and task coordination in the research_mvp local multi-agent runtime (leader, researcher, trainer).
Provides a clear runtime communication contract for the research_mvp local-first multi-agent environment. The skill defines file reading order, CLI commands to inspect threads and inboxes, delegation patterns, and role-specific workflows for leader, researcher, and trainer agents.
Use when operating inside the repository's runtime (tmux-based) to delegate tasks, check inboxes, submit training jobs, perform minimal dry runs, and coordinate cross-agent handoffs. This skill is explicitly tied to the research_mvp runtime and is not intended for general remote or cloud runtimes.
Best suited to agents embedded in or interacting with local runtime tooling (Python-based runtime_cli). Useful for orchestrating human+agent workflows, creating delegation messages, and producing runbooks for experiments.
A well-structured coordination skill for a local multi-agent ML research runtime (leader/researcher/trainer). Clear role separation and detailed workflows for task delegation, training job submission, and result tracking. No scripts included — purely instructional. Very specific to the tinykaggleclaw project's internal runtime, limiting broader applicability.
No security concerns. The skill is project-specific documentation for coordinating agents in an ML research pipeline. Clean, no scripts, no network calls to external hosts. The narrow scope (specific to tinykaggleclaw research_mvp runtime) significantly limits usefulness for other users.
