interview-keyphrase-highlighter

name	interview-keyphrase-highlighter
description	Add or improve Markdown bold emphasis for key phrases in behavioral interview JSON text. Use when highlighting competency briefs, prompts, prompt variants, interviewer intent, evidence, decision criteria, trade-offs, operating mechanisms, red flags, follow-up probes, or rubric signals with **key phrase** markers while preserving valid project data.

Interview Keyphrase Highlighter

Purpose

Add **key phrase** emphasis to natural-language text in data/<group>/<id>/interview.json so generated interview pages are easier to scan. This is an editorial pass: highlight the phrases that carry meaning, not every important-sounding word.

Workflow

1. Read AGENTS.md, the target interview.json, and one nearby interview if useful for local emphasis density.
2. Start with the default first-pass fields: competency brief, prompt, prompt variants, and interviewer intent.
3. Identify the interview's main leadership behaviors, evaluation signals, operating mechanisms, follow-up probes, and weak-answer red flags.
4. Add or adjust **...** only in human-readable text fields.
5. Preserve existing wording unless a tiny edit is needed to make the emphasized phrase grammatical.
6. Keep JSON valid and rebuild generated output after changing source interview data.

Default First Pass

When the user asks to highlight an interview without naming a narrower scope, cover these fields first:

• Competency brief:
  • description
  • competency.definition
  • competency.whyItMatters
• Prompt block:
  • prompt.primary
  • every prompt.variants[] entry
  • every prompt.interviewerIntent[] entry

For prompt text, highlight the behavior being requested, not the generic lead-in. For interviewer intent, highlight the evaluation distinction the interviewer should notice.

Examples:

• Better: influenced a **technical decision without formal authority**
• Weaker: **Tell me about a time** you influenced...
• Better: turn discussion into an **explicit decision mechanism**
• Weaker: Can they **turn discussion**...

Good Phrases To Highlight

Prefer short phrases that help a reader skim senior leadership judgment:

• Competency concepts: **decision criteria**, **customer impact**, **ownership model**.
• Situation constraints: **incomplete data**, **executive ambiguity**, **roadmap pressure**.
• Evidence and mechanisms: **service-level objectives**, **incident taxonomy**, **recurring review**.
• Strong-answer moves: **personal actions**, **measurable results**, **durable operating change**.
• Weak-answer gaps: **no clear baseline**, **generic alignment**, **missing ownership**.
• Rubric signals: **cross-functional alignment**, **portfolio trade-offs**, **decision rights**.

Density Rules

• Use emphasis sparingly: usually one short phrase per sentence or list item.
• Prefer 2-5 word phrases over single generic words or long clauses.
• Do not bold full sentences, whole bullets, headings, or repeated boilerplate.
• Do not bold every occurrence of the same phrase; emphasize where it teaches the reader what to notice.
• If a section is already dense with bold markers, improve the weakest markers instead of adding more.
• Keep ** markers balanced and inside the string value.

Fields To Consider

Natural-language fields commonly worth highlighting:

• description
• competency.definition
• competency.whyItMatters
• prompt.primary, prompt.variants[], prompt.interviewerIntent[]
• scenario.signatureExample, scenario.context[], scenario.constraints[]
• storyAnatomy.situation, task, action[], result[], reflection, systemChange[]
• exampleStories[].situation, strongAnswerSketch[], whyItWorks[], weakVersion, followUpAngles[]
• strongAnswerPattern.summary, strongAnswerPattern.moves[]
• weakAnswerPattern.summary, weakAnswerPattern.redFlags[]
• followUps[].question, followUps[].exampleAnswers.strong[], followUps[].exampleAnswers.weak[]
• evaluation.signals[].strong, evaluation.signals[].weak
• evaluation.rubric[].expected
• practiceTemplate.prompts[].label
• relatedScenarios[]
• toProbeFurther.links[].why

Fields To Avoid

Do not add bold markers to machine-readable or path-like fields:

• id, name, title, family, roleFocus, difficulty, level, source, type, year, url
• asset paths such as assets.icon, explainerComic, aiVisual, strongAiVisual, weakAiVisual
• visualization ids such as visualizations.timeline[], signalMap[], followUpTree[]
• toProbeFurther.links[].id, group, groupDescription, title, url

When unsure, leave the field unmodified.

Editing Guidance

Keep the interview's seniority signal intact. A good highlight points to a real behavior or evaluation distinction:

• Better: created **decision criteria** before execution
• Weaker: created **clarity**
• Better: left behind **ownership and escalation paths**
• Weaker: left behind **important changes**

If highlighting exposes vague prose, make the smallest useful wording improvement and treat the task like a content edit. Coordinate with behavioral-interview-improver when the user asks for broader improvement beyond emphasis.

Verification

After editing interview JSON, run:

python3 -m json.tool data/<group>/<id>/interview.json >/dev/null
python3 _scripts/validate_interviews.py
python3 _scripts/summarize_coverage.py
python3 build.py
python3 _scripts/smoke_static_site.py

For skill-only edits, git diff --check is enough.