twyd-ai/nemeda-agent-kit
Portable methodology, tools, and repository context for AI coding agents.
Write the notes for a transcribed meeting (Summary, Decisions, Action items, Open questions) from a transcript folder produced by nemeda-agent meeting process, file them where the project keeps meeting notes, and log the meeting to the project memory. Use when a transcript exists without notes, when the automatic notes were not generated (no local agent at the time), or when the user wants the notes refined or regenerated.
Log this session's work to the project's memory — decisions made, findings, and a summary segmented by AI model — after explicit user confirmation. Use when the user asks to log the session, record what was done, or write a knowledge-log/memory entry, and the repository has a memory section configured.
Initialize Nemeda Agent Kit safely in a repository by creating a minimal .nemeda/agent-kit.json and AGENTS.md without overwriting existing project guidance. Use when a user asks to adopt, initialize, configure, or migrate a repository to the toolkit.
Load and apply repository-owned Nemeda Agent Kit context before working in a configured repository or multi-repository workspace. Use when a task depends on project identity, repository role, stack profile, required tools, or durable instructions.
Create a complete new project workspace end to end — Drive structure, workspace repository, code repositories, Airtable base, and team onboarding — so every teammate and every AI shares the same context from day one. Use when a user asks to start, create, or scaffold a new project or workspace.
Diagnose whether Nemeda Agent Kit is correctly configured for the current repository, including config validity, instruction paths, repository identity, host availability, and legacy duplication. Use when setup is incomplete, tools or context differ between agents, or a teammate reports the plugin is not working.
Apply the shared Nemeda methodology to cross-repository or client-project work: confirm exact target, preserve dirty checkouts, separate reusable behavior from project configuration, protect secrets, validate proportionally, and publish only when explicitly requested.