| name | behavioral-interview-author |
| description | Create new behavioral interview specs for the spec-driven-behavioral-interviews repository. Use when adding a new interview dataset, choosing category placement, scaffolding data/book or data/examples entries, filling the JSON schema, or preparing a new interview for the static explorer. |
Behavioral Interview Author
Workflow
- Read
AGENTS.md, data/README.md, the target group manifest such as data/book/index.json, and one or two nearby interview.json files in the same competency family.
- Choose the publishable group and category. Use
book for canonical catalog content and examples only for small renderer examples.
- Scaffold the dataset before hand-authoring JSON:
python3 _scripts/scaffold_interview.py --list-families
python3 _scripts/scaffold_interview.py --list-roles
python3 _scripts/scaffold_interview.py book new-interview-id \
--title "New Interview Title" \
--category-id judgment-strategy \
--category-name "Judgment and Strategy"
- Fill the source file at
data/<group>/<id>/interview.json. Do not edit generated files under docs/.
- Preserve the structural answer shape:
Situation -> Task -> Action -> Result -> Reflection -> System Change.
- Rebuild and verify after content changes.
Content Requirements
Include these fields with concrete senior-leadership content:
roleFocus, difficulty, and competency with a clear family and definition.prompt.primary, at least two prompt.variants, and interviewer intent.scenario with a signature example, context, and constraints.storyAnatomy with personal accountability, trade-offs, measurable results, reflection, and durable operating change.- At least two
exampleStories with distinct level, situation, strong answer sketch, why it works, weak version, and follow-up angles.
strongAnswerPattern, weakAnswerPattern, followUps, evaluation.signals, evaluation.rubric, visualizations, practiceTemplate, and relatedScenarios.
Quality Bar
Make the interview test a real senior operating judgment, not a generic behavioral prompt. Strong content names the ambiguity, conflict, constraints, personal decisions, rejected alternatives, measurable outcomes, and mechanisms left behind. Avoid vague claims such as "aligned stakeholders" unless the spec explains the actual forum, decision rule, metric, or ownership model.
Verification
Run the focused checks before finishing:
python3 _scripts/validate_interviews.py
python3 _scripts/summarize_coverage.py
python3 build.py
python3 _scripts/smoke_static_site.py
