rawwerks

A comprehensive collection of Agent Skills for context engineering, multi-agent architectures, and production agent systems. Use when building, optimizing, or debugging agent systems that require effective context management.

This skill should be used when the user asks to "fine-tune on books", "create SFT dataset", "train style model", "extract ePub text", or mentions style transfer, LoRA training, book segmentation, or author voice replication.

This skill should be used when the user asks to "write a post", "check my voice", "look up contact", "prepare for meeting", "weekly review", "track goals", or mentions personal brand, content creation, network management, or voice consistency.

Production-grade techniques for evaluating LLM outputs using LLMs as judges. Use when implementing LLM-as-judge, comparing model outputs, creating evaluation rubrics, mitigating evaluation bias, or building automated quality assessment pipelines.

Transforms external RDF context into agent mental states (beliefs, desires, intentions) using formal BDI ontology patterns. Use when modeling agent mental states, implementing BDI architecture, transforming RDF to beliefs, or building cognitive agents with neuro-symbolic AI integration.

Strategies for compressing context when agent sessions exceed limits. Use when compressing context, summarizing conversation history, implementing compaction, reducing token usage, or optimizing tokens-per-task for long-running agent sessions.

Predictable degradation patterns as context length increases, including lost-in-middle phenomena and context poisoning. Use when diagnosing context problems, fixing lost-in-middle issues, debugging agent failures, or understanding attention patterns and context confusion.

Complete state available to a language model at inference time, including system instructions, tool definitions, retrieved documents, and message history. Use when understanding context, explaining context windows, designing agent architecture, or learning progressive disclosure and context budgeting.

CAD modeling with build123d Python library. Use when creating 3D models, exporting to GLB/STEP/STL, or doing boolean operations (union, difference, intersection). Triggers on: CAD, 3D modeling, sphere, box, cylinder, mesh export, GLB, STEP, STL, solid modeling, parametric design, threads, fasteners, bolts, nuts, screws, gears, pipes, flanges, bearings, bd_warehouse, spur gear, helical gear, bevel gear, planetary gear, ring gear, cycloid gear, rack and pinion, gggears, herringbone, gear mesh, gear train.

Compare 3D CAD models using boolean operations (IoU, Dice, precision/recall). Use when evaluating generated models against gold references, diffing CAD revisions, or computing similarity metrics for ML training. Triggers on: model diff, compare models, IoU, intersection over union, model similarity, CAD comparison, STEP diff, 3D evaluation, gold reference, generated model, precision recall 3D.

Optimize and post-process GLB/glTF 3D models. Use when compressing models for web delivery, reducing file size, simplifying geometry, inspecting model stats, merging models, or converting textures. Triggers on: optimize GLB, compress model, reduce file size, simplify mesh, draco compression, meshopt, webp textures, inspect model, merge GLB, model optimization.

Render GLB 3D models to PNG images for visual verification. Use when you need to SEE your generated 3D model, verify geometry, check visual correctness, or show results to users. Triggers on: render GLB, preview model, visualize 3D, see my model, check geometry, visual verification, screenshot model, model preview.

Recursive Language Models (RLM) CLI - enables LLMs to recursively process large contexts by decomposing inputs and calling themselves over parts. Use for code analysis, diff reviews, codebase exploration. Triggers on "rlm ask", "rlm complete", "rlm search", "rlm index".

Keep a repo's docs/data/code in sync after a change. Use when a change may have left dependent docs or generated views stale, to find exactly what's affected (deterministically) and regenerate/worklist it. Triggers — "true up the repo", "what's stale after this change", "did editing X break the docs", drift between a source-of-truth file and the prose that cites it.

Use when orchestrating multi-step, multi-agent work in Claude Code and one agent turn isn't enough — when a script should hold the loop, the branching, and the intermediate results while your context only sees the final answer. Provides the six canonical dynamic-workflow patterns as runnable scripts: Classify-And-Act, Fanout-And-Synthesize, Adversarial Verification, Generate-And-Filter, Tournament, and Loop Until Done. Trigger on: "fan out / parallelize across N", "verify/refute findings before reporting", "rank or pick the best of several attempts", "sweep until done / find them all", "triage a backlog", "synthesize across sources", or any task that needs a reusable orchestration shape rather than a one-shot prompt.

Autonomous AI coding loop using bd for task management with parallel tmux subagents. Ships features while you sleep.

Send keystrokes to any terminal app via zmx sessions. Use when you need to type into TUI apps, talk to other agents, or drive interactive programs programmatically.

Use this skill when the user wants to call different LLM models through OpenRouter's unified API, compare model responses, track costs and response times, or find the best model for a task. Triggers include requests to test models, benchmark performance, use specific providers (OpenAI, Anthropic, Google, etc.), or optimize for speed/cost.