Labels

Suspected spam or low-quality contribution

Server endpoints, SDK, streaming, A2A/ACP, and public contracts

Browser control, page understanding, and automated web interaction

CLI commands, options, terminal flows, and command-driven setup

Application settings, defaults, environment variables, and feature flags

Dependency additions, removals, upgrades, constraints, and lockfiles

Documentation, guides, examples, architecture, and contributor instructions

UI, chat, dashboard, extension, desktop UI, and accessibility

CI, repository automation, containers, deployment, releases, and packaging

Knowledge ingestion, retrieval, embeddings, and indexing

Licenses, notices, attribution, license policy, and enforcement

Model providers, model selection, inference, and token handling

Agent memory, conversation history, summarization, and retained context

Logs, traces, metrics, activity tracking, and execution trajectories

Agent graph, supervisor routing, delegation, and execution lifecycle

Guardrails, tool approvals, policy matching, and enforcement

LLM prompts, agent instructions, skills, and context construction

Authentication, permissions, secrets, isolation, and sensitive data

Databases, persistence, schemas, migrations, and filesystem behavior

Tests, fixtures, evaluations, benchmarks, scoring, and stability checks

Tool registry, MCP, runtime skills, and external service integrations

Do not merge until evaluation / benchmark validation is done

Something isn't working

Large or risky change — needs careful review

Small, focused change — easy to review

Moderate scope — multiple files or non-trivial logic

Core agent loop, DynamicAgentGraph, LLM node, tool execution, CugaLite

Deployment and infrastructure

UX and visual design

Benchmarks and evaluation