TokenDanceLab
OrganizationIM-native multi-agent collaboration platform — @mention Claude Code, Codex, OpenCode like teammates
Categories
Indexed Skills (7)
adapter-dev
SOP for developing new Agent CLI adapters in AgentHub. Agent should use when adding a new agent integration (e.g. Copilot, Gemini CLI, Qwen CLI) or debugging adapter issues around model resolution, environment, or stream parsing.
dev-loop
Use when AgentHub work is long-running, cross-file, multi-agent, worktree-based, or needs repeated implementation, review, verification, documentation, and progress synchronization.
env-sandbox
Safely manage subprocess environments for AI agent CLIs — whitelist-based env filtering, runtime vars injection, and sensitive key detection.
real-e2e-acceptance
Use when AgentHub work claims real E2E, approved-real, release readiness, packaged Desktop behavior, full-stack data flow, performance acceptance, or before merging branches that change UI workflows, Hub/Edge behavior, runtime execution, login, packaging, or test gates.
integration-test
Use when AgentHub adapter, executor, Edge runtime event pipeline, WebSocket, or local runtime smoke behavior changes.
pre-push
Use when AgentHub changes are committed or nearly ready to push, especially before updating a PR branch or claiming CI-ready status.
test-coverage
Use when AgentHub changes need unit test, coverage, benchmark, or CI coverage-threshold evidence, or when the user asks to check tests or coverage.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.