| name | star-audit |
| description | Audit STAR responses in the candidate's interview-prep documents against his STAR construction conventions before delivery. Use this skill whenever an interview-prep document (recruiter screen, behavioural panel, executive panel) containing STAR responses or a speaking-notes addendum is about to be saved to Interview Prep/ or presented as the final output of a prep session. Also use when [Candidate First Name] says 'audit the STARs', 'check the STAR structure', 'are these STARs right', or 'run the STAR check'. The audit is mechanical (~30–60 sec) and catches the most common prep failure modes: an Action written as a what-list instead of a how-mechanism, numbered-list structure instead of flowing 5.8-style prose, missing or unbolded principle lines, and spoken responses without a drop list of at least 20 seconds of cuttable runway. |
STAR Audit — interview-prep verification skill
Codifies the STAR construction conventions of Project_Instructions.md (rev 2026-06-12). Wave 2 verification skill of the framework.
This skill mechanically audits every STAR response in an interview-prep document. It exists because the STAR disciplines are the most-often-eroded content rules under context pressure: a reshape or tightening pass quietly flattens a HOW into a what-list, or drops the bolded principle lines that are the load-bearing structure [Candidate First Name] holds in memory under panel pressure.
When to invoke
- Default trigger: before saving any interview-prep document or speaking-notes addendum containing STAR responses to
[Workspace Folder Path]/Interview Prep/, and after any content reshape or tightening pass that touched STAR text.
- Explicit triggers: "audit the STARs", "check the STAR structure", "run the STAR check", "are the principle lines intact".
- Skip when: the document is a practicalities-only deliverable (e.g. a short recruiter Q&A with no STAR content).
Inputs the skill needs
- The prep document text (or file path) — the brief, the addendum, or both.
- The interview format (recruiter screen / behavioural panel / executive panel) — determines whether drop lists are mandatory (they are for any spoken response).
Audit checks — run all six per STAR
Load references/star-rules.md for the full rule definitions, then check each STAR response:
- Why/How/What structure. Situation + Task present and framed as why the work was needed; Action present and framed as how (mechanism, sequencing, judgement call); Result present with metrics. Flag any Action that reads as a bare what-list ("did X, did Y, did Z" with no mechanism or judgement).
- 5.8-style prose. The Action is flowing prose, not a numbered list. Each HOW move is followed by an embedded "why this worked" principle. Flag numbered/bulleted Actions as High severity — they read as a checklist, not an executive thinker.
- Bolded principle lines. Embedded principle lines are bolded; the response ends on a crisp closing principle, bolded. Flag missing bolding (Medium) and a missing closing principle (High).
- Drop list present. Every spoken STAR carries an on-the-fly drop list naming specific sentences with per-cut time savings.
- Drop-list runway ≥ 20 seconds. Sum the stated savings; flag below 20 sec (Medium).
- Closing principles never in the drop list. Flag any drop-list entry that names a closing principle line or any bolded principle (High — those are the structural lands the panel remembers).
Output format
A per-STAR table (STAR ID / checks passed / findings with severity and exact-quote evidence / recommended fix), then a one-line verdict: Pass or Amend before delivery (any High finding). Gate behaviour: High findings block delivery; Medium and Low are advisory — [Candidate First Name] decides.
