AI engineering
you can inspect.
Open the full example. Copy the prompt. See what the check actually proves. Practical notes on building and reviewing LLM applications.
20 resources
Structured output validation: 6 invoice checks
Six-case invoice validation lab. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
MCP tool permissions: check user, record and field
Six-case authorization lab. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
Idempotency: test a successful write with a lost reply
Lost-reply simulation. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
LLM evaluation: measure the usable result
Model comparison scorecard. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
Tool calling: test the complete round trip
Tool round-trip test prompt. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
Claude and OpenAI migration checklist
Migration diff prompt and checklist. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
Prompt caching: find the unstable prefix
Prompt layout audit. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
RAG citation verification: inspect the supporting passage
Claim-to-passage review prompt. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
AI agent acceptance checks: prove the task finished
Acceptance and recovery worksheet. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
AI evaluation cases: turn a failure into a regression test
Failure-to-test template. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
LLM context design: resolve conflicting instructions
Context inventory prompt. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
Prompt injection: test the document boundary
Untrusted-document test brief. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
Agent trace review: find the first wrong step
Trace review worksheet. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
Single agent or multiple agents: test the handoff
Architecture decision worksheet. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
Agent approvals: show the exact proposed change
Approval payload template. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
RAG freshness: choose the policy that applies
Policy-version test template. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
LLM streaming latency: measure first token and finish
Streaming measurement sheet. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
Long-running agents: write a useful session handoff
Session handoff prompt. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
AI model releases: turn an announcement into a test
Release-to-test brief. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗
AI engineering portfolio: make the work inspectable
Inspectable case-study template. Includes a worked example, a copyable prompt and a review checklist.
Open the full resource ↗From a short post to a useful working example.
Every resource includes an explanation, a copyable prompt and a checklist. Three notes include deterministic local labs; the others are templates for your own work. Prompts are starting points, and model responses still need review.
Discuss a workflow ↗