Hiring Scorecard

When to use

• Before an interview loop, to give each interviewer a focus, questions and a scale.
• After interviews, to put the written feedback into one evidence-only debrief.
• Not for sourcing candidates, sending offers, or ranking people on traits. The hiring team decides.

Inputs

1. Role requirements. First look for a signed-in tab with the job description or the hiring plan (openbot_browser.list_tabs), or an installed plugin such as Notion (tool_search). Else ask the user to paste the job description and the level.
2. Real work. Two or three tasks the person will do in the first 90 days. If the user gives none, ask once with ask_user. Competencies come from this work, not from the job title.
3. Interviewers. Names and the stage each one runs. If missing, write the plan with "Interviewer 1, 2, 3".
4. Feedback (for a debrief). Scorecards or notes pasted by the user or read from a signed-in tab.

Do not store candidate names, notes or scores in shared data tables. Other agents can read them. Keep them in the chat.

Steps

1. Competencies. Write 4 to 6 competencies. Each one is a work behavior that someone can observe, such as "breaks an unclear request into a plan with checkpoints". Tie each one to a requirement or a task.
2. Coverage. Give each interviewer 1 or 2 competencies. Each competency is covered by at least one interviewer and by at most two. Show the coverage as a table.
3. Questions. For each competency, write one primary question and two follow-up probes. Use questions about past work ("Tell me about a time you ...") or a short work sample. Every candidate gets the same primary questions.
4. Anchors. For each competency, write one behavior for each score on the 1 to 4 scale (see the rubric).
5. Prep pack. For each interviewer: their competencies, questions, anchors, time plan, and one line on what not to ask (see stop conditions).
6. Debrief. Ask each interviewer for scores and evidence before they read other feedback. Then build one table: competency, score per interviewer, quoted evidence. Mark a score without evidence as "no evidence". List disagreements of 2 or more points. List the competencies with no evidence and the follow-up question that would close the gap.
7. Chase. List interviewers whose scorecard is missing and draft a short reminder for the user to send.

Rubric

A 1 to 4 scale. Write one behavioral anchor for each score, for each competency.

• 1 — no evidence or opposite evidence. Example: "could not name a decision they made on the project".
• 2 — partial. Did the work with heavy guidance, or only described the team's work.
• 3 — meets the level. Did the work alone, gave a specific example with the result.
• 4 — above the level. Did the work and improved how others do it, with a measured result.

Rules: a score needs a quote or a concrete note from the interview. "Good culture fit", "seemed smart" and "strong presence" are impressions, not evidence; list them apart and give them no weight.

Output

Role: <title, level> — loop of <n> interviews

Competencies and coverage
| Competency | From requirement | Interviewer |

Prep pack — <interviewer>
Focus: <competency 1>, <competency 2>
Primary question: ...  Probes: ...
Anchors: 1 ... 2 ... 3 ... 4 ...
Do not ask: age, family, health, religion, nationality, or plans outside work.

Debrief — <candidate reference>
| Competency | <Interviewer A> | <Interviewer B> | Evidence |
Disagreements: ...
No evidence: <competency> — next question: ...
Missing scorecards: <names>
Decision: for the hiring team.

Approvals and stop conditions

• Do not email, message or schedule candidates or interviewers without the user's yes. Drafts stay in the chat.
• Do not change records in an applicant tracking system unless the user asks for that edit.
• Never use or infer protected characteristics: age, sex, gender identity, race, ethnicity, religion, disability, health, pregnancy, family status, sexual orientation, nationality. Stop and flag feedback that mentions them.
• Do not give a hire or no-hire recommendation. Show the evidence; the team decides.

Common failures

• Competencies copied from the job title, not from the work.
• Two interviewers asking the same question while one competency gets no coverage.
• Averaging scores that have no evidence behind them.
• Letting the first opinion in the debrief set the scores of the rest.
• Personal traits ("energetic", "polished") written as job evidence.