V Vibe Coding Wiki
Overview 0. Welcome & How to Use This Guide0.1 What "vibe coding" means (and what it doesn't)0.2 Who this guide is for (beginner → expert tracks)0.3 How to follow along (projects, checkpoints, and "ship points")0.4 Required vs optional tools (AI Studio, editor, Git, runtime)0.5 A note on UI changes & product naming
Overview
Overview 1.1 What the model is actually doing1.2 The difference between knowledge, reasoning, and tools1.3 Context windows and why "just add more text" fails1.4 Temperature, randomness, and creativity knobs1.5 Hallucinations: the predictable failure mode
Overview 2.1 The vibe loop (idea → prompt → output → test → refine)2.2 When vibe coding is great vs when it's dangerous2.3 "You drive, the model types" as the default posture2.4 The 80/20 rule: get to runnable, then get to correct2.5 The "spec first" habit that prevents spaghetti
Overview
Overview 3.1 AI Studio vs Google Cloud/Vertex tooling (conceptual map)3.2 The prompt playground concept3.3 Common workflows: chat, structured output, tool calling, multimodal3.4 The "export code" mindset
Overview 4.1 Account access and permissions basics4.2 API keys vs project credentials (the safe mental model)4.3 Rate limits and quotas (why your prototype suddenly fails)4.4 Basic safety settings and why they matter
Overview 5.1 A one-prompt micro app: "build me a CLI calculator"5.2 Turn the output into a runnable project5.3 The first refactor prompt (make it clean, not clever)5.4 Add one feature without regressions5.5 Capture a reusable template
Overview
Overview 6.1 Ask for a plan before code6.2 Define constraints: language, runtime, libraries, and style6.3 Define "done" with acceptance criteria6.4 Use examples as "mini tests"6.5 Use a "stop and confirm assumptions" step
Overview 7.1 The "scaffold → implement → refine" pattern7.2 The "small modules, stitched later" pattern7.3 The "write tests first" vibe pattern7.4 The "explain then rewrite" pattern7.5 The "diff-only changes" pattern
Overview 8.1 How to paste errors so the model actually helps8.2 The "hypothesize → test → iterate" loop8.3 When the model suggests nonsense: how to recover fast8.4 Prompting for logging and observability8.5 "Fix it without breaking it": regression discipline
Overview
Overview 9.1 Choosing a model for prototypes9.2 Choosing a model for complex refactors9.3 When to switch models mid-project9.4 Controlling determinism for repeatable builds9.5 Cost intuition: tokens, context size, and repeated calls
Overview 10.1 System vs developer vs user instructions (practical rules)10.2 Stable "house rules" prompts for coding projects10.3 Using checklists inside prompts10.4 Separating "spec" from "conversation"
Overview 11.1 What to include vs exclude from context11.2 Summarize state between iterations11.3 The "context budget" technique11.4 Handling multi-file projects without losing coherence11.5 Preventing conflicting instructions over time
Overview 12.1 What safety filters can and cannot do12.2 Designing prompts that avoid risky behavior12.3 Building "refusal-aware" UX in your app12.4 Handling sensitive data responsibly12.5 Audit trails: saving prompts and outputs safely
Overview
Overview 13.1 Turning a prototype into a repo13.2 Environment setup: local dev, env vars, secrets13.3 Basic request/response wrapper architecture13.4 Logging and error handling patterns for LLM calls13.5 Versioning prompts (treat prompts like code)
Overview 14.1 Spec: "Summarize any article into structured bullets"14.2 UI option A: CLI tool14.3 UI option B: small web app14.4 Data flow: input → model → structured output → display14.5 Hardening: timeouts, retries, and fallbacks
Overview 15.1 JSON-first prompting15.2 Designing schemas that are hard to break15.3 Handling partial/invalid JSON gracefully15.4 Using enums and constraints to reduce ambiguity15.5 Validating outputs and surfacing errors to users
Overview 16.1 What "tools" are (and why they reduce hallucination)16.2 Designing tool interfaces (inputs/outputs) cleanly16.3 The "planner-executor" vibe pattern16.4 Tool error handling (when APIs fail)16.5 Preventing tool abuse and runaway loops
Overview
Overview 17.1 Idea → one-page spec in 10 minutes17.2 Architecture sketching with the model17.3 Generating scaffolding: folders, configs, and scripts17.4 Building the "walking skeleton"17.5 Iterating features with controlled scope
Overview 18.1 "Repo tour" prompts that work18.2 Creating a system-level map (modules, dependencies, entrypoints)18.3 Safe refactoring prompts (small steps, diffs, tests)18.4 Legacy code: extract seams before changing behavior18.5 Migrating frameworks or libraries with AI help
Overview 19.1 Turning logs into hypotheses19.2 Reproducing bugs with minimal test cases19.3 Fixing with guardrails: tests, assertions, and contracts19.4 Postmortems: writing a useful incident report19.5 Preventing recurrence: monitoring and alerts
Overview 20.1 Generate docs from code structure20.2 "Docs as a build artifact" workflow20.3 API docs: examples, edge cases, and error contracts20.4 README that gets people to "hello world" fast20.5 Changelogs and migration notes with AI assistance
Overview
Overview 21.1 Image inputs for debugging UI issues21.2 Extracting structured data from images (carefully)21.3 Image-based UX critique prompts21.4 Visual test case generation21.5 Safety and privacy with user images
Overview 22.1 Chunking strategies that preserve meaning22.2 Building a "document Q&A" assistant prototype22.3 Citation-like behavior (trace outputs to sources)22.4 Handling contradictions and multiple versions22.5 Long-context performance and cost tradeoffs
Overview 23.1 Turning meetings into action items and tickets23.2 Generating transcripts and cleaning them23.3 Extracting decisions vs discussion23.4 Building a searchable "knowledge log"23.5 Guardrails for sensitive recordings
Overview
Overview 24.1 When you need RAG (and when you don't)24.2 Choosing chunk size, overlap, and metadata24.3 Embeddings 101 for builders24.4 Ranking and re-ranking intuition24.5 Prompting the model to use retrieved context faithfully
Overview 25.1 Spec: "Answer questions about my docs with references"25.2 Indexing pipeline: ingest → chunk → embed → store25.3 Query pipeline: retrieve → compose prompt → answer25.4 Evaluation: measuring answer quality and faithfulness25.5 Maintenance: updating the index and handling deletions
Overview 26.1 "Answer only from sources" prompting patterns26.2 Confidence & uncertainty UX26.3 Conflict detection in retrieved sources26.4 Refusal and escalation flows26.5 Auditing: storing which chunks influenced answers
Overview
Overview 27.1 What you can unit test vs what you must evaluate statistically27.2 Golden tests for structured outputs27.3 Property-based tests for robustness27.4 Fuzzing prompts and inputs27.5 Snapshot testing with careful update workflows
Overview 28.1 Build a tiny eval set that matters28.2 Scoring outputs with rubrics28.3 Pairwise comparisons for model/prompt tuning28.4 Regression detection across prompt versions28.5 Human review workflows that don't waste time
Overview 29.1 Timeouts, retries, and idempotency29.2 Circuit breakers and fallback modes29.3 Streaming responses and partial rendering29.4 Caching strategies (prompt+context caching)29.5 Observability: traces, metrics, and prompt logs
Overview
Overview 30.1 What attackers want from your AI app30.2 Data exfiltration risks (secrets, PII, proprietary data)30.3 Indirect prompt injection (documents as attackers)30.4 Tool misuse (dangerous function calls)30.5 Supply chain risks (dependencies + generated code)
Overview 31.1 Input sanitization and allowlists31.2 Output filtering and schema enforcement31.3 Least-privilege tool design31.4 Secrets handling: never in prompts, never in logs31.5 Red-teaming your own prompts
Overview
Overview 32.1 Why tokens cost more than you think32.2 Summarize and compress context safely32.3 Choosing small vs large models strategically32.4 Batch processing vs interactive mode32.5 Measuring cost per successful task
Overview 33.1 Cutting prompt size without losing accuracy33.2 Streaming UX patterns that feel instant33.3 Parallelizing retrieval and preprocessing33.4 Caching: what to cache and what not to33.5 Warm starts and connection reuse
Overview
Overview 34.1 Requirements: what it should and shouldn't do34.2 Repo indexing strategy34.3 Change proposal workflow (plan → diff → tests)34.4 PR descriptions and code review assistance34.5 Guardrails: refuse risky changes and require approvals
Overview 35.1 Intent classification and routing35.2 Structured extraction (order IDs, dates, issue types)35.3 Suggested replies with policy constraints35.4 Escalation logic and confidence thresholds35.5 Evaluation and privacy controls
Overview 36.1 Upload/ingest flow for CSV-like data36.2 Ask questions in natural language; answer in charts/tables36.3 Schema inference and validation36.4 Reproducible "analysis notebooks" from AI output36.5 Guardrails: avoid confident math errors
Overview
Overview 37.1 What an "agent" is in practice37.2 Single-agent vs multi-agent designs37.3 Planning, executing, and critiquing roles37.4 Termination conditions and budgets37.5 Auditability and replay
Overview 38.1 The "proposal-only" agent pattern38.2 The "diff generation" agent pattern38.3 Test-driven agent behavior38.4 Static analysis and lint gates38.5 Security gates and dependency scanning
Overview 39.1 Hierarchical prompts: global rules → module rules → task rules39.2 Prompt "APIs": stable interfaces between app and model39.3 Self-critique patterns (and when they backfire)39.4 Multi-step reasoning with enforced intermediate artifacts39.5 Prompt compression and distillation
Overview 40.1 Deep schemas: nested objects and arrays40.2 "Explain errors as JSON" for debuggability40.3 Schema evolution and backwards compatibility40.4 Internationalization schemas (multi-locale outputs)40.5 Contract tests for schema compliance
Overview 41.1 Hybrid search (keyword + vector)41.2 Metadata filtering and access control41.3 Query rewriting and expansion41.4 Context packing: selecting the best chunks under budget41.5 "Answer with provenance" techniques
Overview 42.1 When fine-tuning helps (style, format, narrow tasks)42.2 When fine-tuning hurts (rapidly changing knowledge)42.3 Data requirements and dataset quality42.4 Evaluation before and after tuning42.5 Deployment and rollback strategy
Overview
Overview 43.1 Prompt reviews like code reviews43.2 Shared prompt libraries and templates43.3 "Spec → prompt → eval → ship" as a team pipeline43.4 Onboarding: teaching new devs the vibe loop43.5 Defining quality bars for AI features
Overview 44.1 Data classification and what can enter prompts44.2 Logging policies: what to store and redact44.3 User consent and transparency UX44.4 Incident handling for AI mistakes44.5 Model/provider change management
Overview
Overview 45.1 The model ignores instructions45.2 Outputs are correct but messy45.3 Outputs are confident but wrong45.4 It works once but not reliably45.5 It's too slow/expensive
Overview 46.1 Pre-build checklist (spec, constraints, acceptance tests)46.2 Prompt checklist (inputs, outputs, edge cases, failure paths)46.3 Tool checklist (permissions, budgets, timeouts)46.4 Production checklist (logging, monitoring, rollbacks)46.5 Security checklist (prompt injection, secrets, access control)
Overview 47.1 The universal coding system prompt47.2 The "plan first" prompt47.3 The "diff-only" prompt47.4 The "write tests then implement" prompt47.5 The "explain tradeoffs" prompt
Overview 48.1 Key terms: tokens, context, temperature, tools, RAG48.2 Common schema patterns48.3 Error patterns and what they usually mean48.4 Architecture patterns for LLM apps48.5 A "choose your next step" map
Home/ Part XV — Troubleshooting, Checklists, and Reference/46. Vibe Coding Checklists

46. Vibe Coding Checklists

Overview and links for this section of the guide.

On this page

  • Pre-Flight Checks
  • The Lists
  • Where to go next

Pre-Flight Checks

Pilots use checklists because memory is fallible. AI Engineers should too.

The Lists

  • Pre-Build: Do we actually need AI for this?
  • Prompting: Did we handle edge cases?
  • Security: Is the API key safe?

Where to go next

  • 46.1 Pre-build checklist
  • 46.2 Prompt checklist
  • 46.3 Tool checklist
  • 46.4 Production checklist
  • 46.5 Security checklist

Explore next

46. Vibe Coding Checklists sub-sections

5 pages

46.1 Pre-build checklist (spec, constraints, acceptance tests)

Open page

→

46.2 Prompt checklist (inputs, outputs, edge cases, failure paths)

Open page

→

46.3 Tool checklist (permissions, budgets, timeouts)

Open page

→

46.4 Production checklist (logging, monitoring, rollbacks)

Open page

→

46.5 Security checklist (prompt injection, secrets, access control)

Open page

→
← Previous 45.5 It's too slow/expensive Next → 46.1 Pre-build checklist (spec, constraints, acceptance tests)

On this page

  • Pre-Flight Checks
  • The Lists
  • Where to go next

Vibe Coding Wiki — Build faster with AI assistance