Terms in this guide: Evaluation · Fixture · Verification
SKILL LAB / RECORDED PRACTICE
A book-club request with missing details tests whether a brief can stay useful without inventing facts.
What we evaluated
The task was to prepare documentation for a local book-club event page. The request named the audience and goal but left the meeting date, venue, book title, and registration decision unresolved. The supplied folder contained no website, framework, hosting details, budget, or deadline. The assistant read the published project-brief skill and manually followed its procedure.
The result
The resulting brief identified the visitor’s goal, constrained the work to documentation, listed the unresolved inputs, and stated observable acceptance criteria. It explicitly deferred page construction and registration work. No event facts or technical dependencies were invented.
# Book-club event page brief
## Outcome
Help local book-club visitors find the next meeting date, venue, and book.
## Facts
This exercise contains documentation only. No website or framework exists here.
## Scope
Prepare this brief and an implementation checklist. Do not build a page or registration form.
## Open questions
Confirm meeting date, venue, book title, and whether registration is needed. Confirm design, hosting, and framework before implementation. No deadline or budget is specified.
## Acceptance criteria
The reviewer can identify every required event fact and every missing input. No invented event details appear. Documentation changes only. Before future page construction, obtain the unresolved information and confirm scope.
## Future verification
When a page is authorized, check that confirmed event facts are readable on mobile and desktop. Registration remains outside scope unless separately approved.
How the review was checked
Six local checklist assertions confirmed the outcome, documentation boundary, open questions, absence of an invented dependency, acceptance section, and review gate. The assistant also compared the brief with the full request and inventory. These assertions confirm properties of this output; they do not measure general model quality.
Try the same task
Download the beginner practice project. Read REQUEST.md and INVENTORY.md, draft your own BRIEF.md, and compare it with ACCEPTANCE.md before opening the sample solution. A different brief can be valid if it preserves the same facts and boundaries.
Limits of this evaluation
One assistant-authored documentation exercise was evaluated on September 10, 2026. There was no independent evaluator, automatic skill-selection test, production implementation, or repeated model-run comparison. Treat this as a transparent worked example with a narrow scope.
Use the skill and the exercise
Progress stays in this browser. No account needed.