Complete corpus
Browse the tutorial corpus
This index preserves the full tutorial source set while keeping the short marketing-oriented tutorial landing page discoverable for new visitors.
OverviewLLM Development Skills, Open Horizons AppliedThis repo teaches practical LLM-based development in existing systems: how to shape the task, assemble a prompt with the right context, choose the right solution level, verify the work, and preserve what the neGuideDocs HomeStart here when the repo feels like a pile of good pages but you need a path.GuideCurriculumThis curriculum is for builders using LLMs inside existing systems, where there are real constraints, real tests, real maintenance risk, and more than one plausible answer.GuideBuilder Loop TutorialUse this tutorial for one real project slice: clarify the aim, frame the problem, choose a solution level, define evidence, delegate the work, review it, and preserve what the next session needs.GuideContext to Agent TutorialUse this tutorial when a one-off prompt starts becoming an interface the next session should reuse.GuideEval TutorialUse this tutorial when “the agent passed” is not enough evidence.GuideArtifact ContractsArtifacts are not note-shaped souvenirs. Each artifact must let the next consumer continue the work without rediscovering the conversation.GuideIntent EngineeringIntent Engineering is the first lesson because it names the failure mode.GuideModel-Fit FramingModel-fit framing shapes the task so the model uses its strengths: extracting structure, comparing options, rewriting for an audience, critiquing against criteria, or translating supplied context into a checkabGuideContext ConstructionContext construction is the skill of deciding what the agent should know, what it should not assume, and what should survive the handoff.GuidePrompt and Context AssemblyHere, prompt engineering means interface design: assemble instructions, data, examples, output shape, and checks so the model can do the work over supplied reality.GuideWhat Open Horizons AddsOpen Horizons keeps an agent run from becoming “ask for patch, accept patch.”GuideProblem SpaceProblem space maps the terrain before the agent chooses a fix.GuideProblem StatementA problem statement narrows mapped terrain into the slice the agent is allowed to solve.GuideBeyond the Nearest Peak, Applied to Coding AgentsAgents make it cheap to generate patches. They do not make it cheap to know which patch deserves to survive.GuideEvidence and EvalsEvidence is the boundary between useful agent work and polished guessing.GuideAgent BriefsAn agent brief turns a selected solution into bounded execution.GuideStrategy Clarity for Agent-Driven Project ImprovementAn agent brief is a small strategy doc.GuideAuthoring SkillsA skill is how a repeated procedure survives the chat.GuideAuthoring SubagentsA subagent is how a role survives the chat.GuideExecution, Review, Dissent, and SalvageExecution is the inner loop. Review closes it. Dissent stress-tests it. Salvage preserves learning when the run drifts.GuideKnowledge ExtractionKnowledge extraction is how the session stops being a one-off.GuideFurther ReadingRead these when a module is the bottleneck. The curriculum is overview first, then deep dive, then references.ExampleExample: Duplicate Notification Improvement BriefThis is a worked example for `docs/tutorial.md`.TemplatesModel-Fit Note TemplateUse this before asking an LLM for output.TemplatesContext Pack TemplateBuild this before asking an agent to implement.TemplatesPrompt Assembly TemplateUse this after the intent note, model-fit note, and context pack. The goal is not to fill every field; it is to make each included field earn its place.TemplatesProblem Statement TemplateUse this after the problem-space map and before solution search.TemplatesEvidence and Eval ChecklistDefine the evidence before asking the agent to implement.TemplatesAgent BriefWhat project improvement should this run make?TemplatesProject Improvement PlaygroundUse this to choose a real project slice.TemplatesProject Skill TemplateCopy this into:TemplatesSubagent TemplateCopy this into:TemplatesKnowledge Artifact TemplateChoose the smallest artifact that changes future behavior.