Workflow guide for porting a PyTorch `timm` model to a numerically-equivalent Lux.jl implementation. Load this skill whenever the task involves porting, converting, or translating a PyTorch model (especially anything from `timm.create_model`, `forward_features`, ResNet/ViT/EfficientNet/ConvNeXt/etc. backbones) to Lux; writing or editing Lux `@compact` blocks that must match a PyTorch reference; producing or consuming HDF5 parity fixtures, `read_parity`, `apply_state_dict`, or `@test isapprox` parity tests; loading `.safetensors` weights from HuggingFace Hub in a Julia context; or reasoning about PyTorch-vs-Lux numeric differences (cross-correlation vs convolution, padding semantics, GroupNorm/BatchNorm defaults, weight standardization, NCHW vs WHCN). This skill layers on top of `kaimon-julia`, which remains the source of truth for driving the Julia REPL.
Bootstrap guide for building a Tachikoma.jl TUI app. Load when asked to create, scaffold, or extend a terminal UI in Julia using Tachikoma, including any work that touches Model/update!/view, the app() loop, layouts, widgets, or KeyEvent / MouseEvent handling. Covers the minimal app skeleton and the sharp edges that cost the most round trips.
Workflow guide for driving Julia projects through the Kaimon MCP server. Load this skill whenever working with Julia code in any capacity, including reading, writing, editing, or debugging .jl files, Project.toml/Manifest.toml, Julia REPL sessions, Pkg operations, Revise-based hot-reload workflows, or any Kaimon tool calls (start_session, ex, manage_repl, pkg_add, pkg_rm, check_eval, debug_exfiltrate, goto_definition, workspace_symbols). Kaimon is the preferred execution path for all Julia work in this repo.