🎯Skills3
🔌Plugins10
Execute Python scripts on remote GPU/TPU clusters via SkyPilot
Plan and render GKE TPU Jobs and nodepool actions with explicit context/namespace safety
TPU/XLA profiling analysis: MCP tools for XProf HTTP API (operator breakdowns, memory profiles, A/B comparisons) + methodology skills for trace parsing, MFU calculation, bottleneck identification
Open, list, and inspect Google Cloud support cases (Cloud TPU / GKE / GCS) via the Cloud Support API v2, with guided classification, priority confirmation, and formatted case detail output
Records the complete session content and logs it to a daily work directory with a dynamic filename based on the active CLI agent. Use this for automated progress tracking and documentation.
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
Systematic analysis of TPU pretraining efficiency. profile-anatomy: schema + reference scripts for xplane.pb / trace.json.gz. comm-analysis: per-axis NCCL bus BW, peak ICI utilization, and per-step compute/comm overlap. compute-breakdown: source-line aggregation, layer scoping, non-compute audit, v7x roofline shortfall. memory-profile: HBM peak occupancy, alive-buffer attribution, lifetime/shape/tf_op/parent_jit rollups.
Beaver project management: GitHub-native issue lifecycle with explicit /beaver-xxx commands covering the full development cycle (create, design, decompose, dev, PR, beaver-fix, tracker, focus, setup)
Mine local Claude/Codex session history to produce structured daily/weekly work recaps with optional GitHub Issue sync
Check and fix lint issues for changed Python files by commit, range, or working tree