| name | pace-audit |
| description | Run an honest timing audit on the spoken responses in the candidate's interview-prep documents, at the candidate's actual measured speaking pace — never at the generic 165 wpm executive baseline. Use this skill whenever a speaking-notes addendum or prep document with spoken responses is about to be delivered, AND after every content reshape or tightening pass (timing drifts with every edit). Also use when [Candidate First Name] asks 'how long does this run', 'time these responses', 'am I over budget', or 'run the pace check'. The audit is mechanical (~10–30 sec): word count ÷ actual wpm vs the format's time budget, with a two-pass tightening plan for any response that runs over. |
Pace Audit — interview-prep timing verification skill
Codifies Convention 11 (honest timing audit) and Convention 9 (two-round tightening) of Project_Instructions.md (rev 2026-06-12). Wave 2 verification skill of the framework.
A response of 525 words runs ~3:11 at 165 wpm but ~3:37 at the candidate's actual ~145 wpm — and the gap matters when the budget is tight. This skill makes the timing reality-check mechanical so it cannot silently default to the flattering baseline.
When to invoke
- Default triggers: before delivering any speaking-notes addendum or prep document containing spoken responses; after every reshape or tightening pass (Convention 11: re-audit after every reshape).
- Explicit triggers: "time these", "how long does this run", "am I over", "run the pace check".
- At panel-prep kickoff: confirm the candidate's current pace once per session if not already stated — use the pace measured at setup.
Inputs the skill needs
- The spoken responses (text or file path) — typically Section A tailored responses and Section B backup STARs of an addendum.
- The time budget per response — from recruiter/panel intel if available, else format defaults: ~90 sec recruiter screen; 2–3 min behavioural panel; 2–3 min executive panel.
- the candidate's confirmed pace for this session (per the setup measurement; markedly lower effective rate where [pause] markers are dense).
Method
Load references/pace-method.md, then per response:
- Count words (exclude stage directions:
[pause], [STOP — …], drop-list tables, headings).
- Compute runtime at the confirmed wpm; add ~2–3 sec per
[pause], ~1–2 sec per [short pause].
- Compare to budget. Verdict per response: Under / On / Over (Over = >10% past budget).
- For every Over response, produce the two-pass tightening plan: pass 1 = the existing drop-list items (~15–20 sec recovery); pass 2 = structural cuts inside paragraphs (compressed result-block listings, one of two parallel-structure sentences,
[pause] markers between sub-sections) for ~25–35 sec total. Never cut closing principle lines.
Output format
A timing reality-check table (Response / words / runtime at actual wpm / budget / verdict / recovery needed), then a one-line overall verdict. Gate behaviour: any response >25% over budget blocks delivery until a tightening pass runs; 10–25% over is advisory with the two-pass plan attached.
