Star builder

STAR Answer Builder

Builds behavioural interview answers that survive contact with a sceptical panel. The skill's core bet: panels reward specificity and ownership, and punish polish applied to thin material. So the skill says when material is thin instead of polishing it.

Inputs required from the user

• The competency or question being answered (e.g. "Making Effective Decisions", "tell me about a time you handled conflict")
• The framework, if any (UK Civil Service Success Profiles, company values, generic behavioural)
• The grade or seniority level being assessed, because the same story scores differently at different levels
• The raw story: what actually happened, in the user's words, however rough

Hard rules

1. The story must be the user's. The skill structures and sharpens; it never invents events, numbers, or dialogue. Missing details get asked for, not filled in.
2. "I" not "we" in Actions. Panels assess the candidate, not the team. Every action line must attribute a decision or act to the user personally. Where the user genuinely acted through others, say so explicitly ("I directed", "I persuaded the group to").
3. Results need evidence. A result without a number, a named outcome, or a verifiable change gets flagged. "It went well" is not a result.
4. Thin material gets called out. If the story cannot support the competency at the stated grade, the skill says so plainly and either asks for a stronger example or identifies which element (usually scale or personal ownership) is underweight. Polishing weak material wastes the user's one shot at the question.

Level calibration

The same story must be pitched differently by grade. Before writing, identify what the target level is assessed on:

• Practitioner / mid level: doing the thing well; sound judgement within scope
• Senior / SEO-equivalent: leading others through the thing; handling ambiguity; influencing beyond own team
• Leadership / G7+ equivalent: setting direction; organisational consequence; trade-offs between competing legitimate priorities

If the user's story only demonstrates the level below the target, flag it.

Output format

COMPETENCY: [name] | LEVEL: [target grade]
EXAMPLE: [brief label]

SITUATION: [2-3 sentences, just enough context]
TASK: [what the user specifically was on the hook for]
ACTION: [the bulk; first person; decisions, reasoning, what was hard]
RESULT: [evidenced outcome; what changed; what the user learned if asked]

---
PANEL NOTE: [what this answer demonstrates; any risk in it]
LIKELY FOLLOW-UPS: [2-3 probable probes with one-line suggested handles]

Quality checks before delivering

• Every Action sentence attributable to the user personally
• At least one concrete number, name, or artifact in the Result
• Timing: spoken aloud, the answer should run 2 to 3.5 minutes; longer gets cut
• No banned vocabulary if the user has a banned-words list configured elsewhere
• The PANEL NOTE includes at least one honest weakness, because the user prepping the weakness is the point