Job Description And Scorecard Builder

Job Description and Scorecard Builder

This skill generates a job description (JD) and a matched interview scorecard from your role input notes in a single pass. Designed for hiring managers, team leads, and department heads who run interviews without a dedicated recruiter. No recruiting SaaS required.

Input:

• Role notes (free-form text): title, level, must-have requirements, nice-to-have, compensation range, work format, optional team context

Output:

• Single markdown response with two paired documents: Job Description + Interview Scorecard, ready to copy-paste

---

Language Detection

Detect the user's language from their message:

• If Russian (or contains Cyrillic): respond in Russian
• If English (or other Latin-script language): respond in English
• If ambiguous: respond in the language of the trigger phrase used

---

Instructions

Step 1: Parse Role Input

1. Read the user's role notes (free-form text, bullet list, or structured).
2. Extract structured fields:

• Title: role name (required)
• Level: Junior / Middle / Senior / Lead / Director / VP / C-level
• Must-have requirements: list of non-negotiable skills or qualifications
• Nice-to-have requirements: list of preferred but optional skills
• Compensation: range or "not specified"
• Work format: Remote / Hybrid / Onsite
• Team context: team size, reporting line, or department (optional)
• Interview format preference: competency-based / case-based / behavioral (optional)

1. If Title is present but other fields are missing: proceed with available data; flag gaps at the top of the output.
2. If input is only a job title with no other context (e.g., "Engineer"):

• Generate 4–5 clarifying questions (level, must-haves, team context, compensation, work format)
• Do not produce an empty template
• Respond: "Роль пока не описана — нужно больше деталей. / The role is underspecified. Please answer these questions:"
• Stop. Do not proceed to document generation.

Step 2: Generate Job Description (Format A)

1. Create the JD with these sections in order:

• About the Role: 1–2 sentences on what the person does and team context; include why the role exists if known
• Responsibilities: 4–7 bullet points, each starting with an action verb (Build, Lead, Drive, Analyze, etc.)
• Requirements > Must Have: derived from must-have input; one requirement per bullet
• Requirements > Nice to Have: derived from nice-to-have input; omit section entirely if none provided
• Compensation & Work Format: compensation range + work format on separate bullets
• About the Team: 1 sentence on team size, structure, or reporting line; omit if no team context provided

1. Adjust language register:

• Standard roles (Junior through Lead): professional, specific, results-oriented
• Leadership roles (Director / VP / C-level): strategic language, focus on organizational impact

Edge Case — Leadership Role: If level is VP, Director, or C-level: add leadership, strategic thinking, and organizational impact to the Responsibilities list; these same dimensions become criteria in the Scorecard (Step 3).

Edge Case — Update Mode: If user pastes an existing JD with a request to update it: apply new inputs as changes only; prefix output: "Updated: [list of changed sections]. Unchanged sections kept as-is." Do not regenerate the full JD from scratch.

Edge Case — Mixed Language Input: If notes are in mixed EN + RU: detect dominant language; generate both documents in that language; translate all requirements consistently.

Step 3: Generate Interview Scorecard (Format B)

1. Map each must-have requirement from the JD to a scored criterion in the scorecard. Criterion names must match the JD requirement phrasing exactly (preserves the paired linkage).
2. Assign weights:

• Must-have requirements → High
• Nice-to-have requirements → Medium (add at the end of the criteria table; omit if none provided)

1. For each criterion, write behavioral anchors:

• 1 — Not demonstrated: no evidence observed in the interview
• 2 — Below bar: partial or inconsistent evidence; lacks depth or breadth required for the role
• 3 — Meets bar: clear evidence meeting the stated JD requirement
• 4 — Exceeds: significantly stronger; demonstrates expertise beyond what the role requires

1. Add Overall Recommendation with four options: Strong Hire / Hire / No Hire / Strong No Hire
2. Add Notes / Observations free-text field at the end

Edge Case — Interview Format Specified: If user specified a format (case / behavioral / technical): align anchor language to that format; add a note at the top of the scorecard: "Scorecard designed for [format] interview."

Step 4: Format and Output

1. Output both documents in a single markdown response:

• Format A (JD) first, labeled ## Job Description — [Title] ([Level])
• Horizontal rule (---) as separator
• Format B (Scorecard) second, labeled ## Interview Scorecard — [Title] ([Level])

1. No meta-commentary between or after the two documents
2. Output language matches detected user language
3. All section headers translate to Russian when responding in Russian:

• "About the Role" → "О роли"
• "Responsibilities" → "Обязанности"
• "Must Have" → "Обязательно"
• "Nice to Have" → "Желательно"
• "Compensation & Work Format" → "Компенсация и формат работы"
• "About the Team" → "О команде"
• "Interview Scorecard" → "Оценочный лист"
• "Overall Recommendation" → "Итоговая рекомендация"
• "Notes / Observations" → "Заметки / Наблюдения"

---

Negative Cases

• Empty input or whitespace only: Respond: "Опиши роль: название, уровень, ключевые требования. / Describe the role: title, level, must-have requirements."
• Request to find or search candidates: Respond: "Этот скилл создаёт документы — вакансию и скоркард. Поиск кандидатов не поддерживается. / This skill creates documents — a job description and scorecard. Candidate search is not supported."
• Title only, no other context: Ask clarifying questions (see Step 1.4). Do not produce an empty template.

---

Output Format

Single markdown response with two paired, labeled sections:

## Job Description — [Role Title] ([Level])

### About the Role
[1–2 sentences: what the person does, team context, why the role exists]

### Responsibilities
- [Action-verb-first responsibility]
- [Action-verb-first responsibility]
- [Action-verb-first responsibility]
- [Action-verb-first responsibility]

### Requirements

**Must Have:**
- [Must-have requirement 1]
- [Must-have requirement 2]

**Nice to Have:**
- [Optional requirement 1]

### Compensation & Work Format
- **Compensation:** [range or "not specified"]
- **Work format:** [Remote / Hybrid / Onsite]

### About the Team
[1 sentence: team size, structure, or reporting line]

---

## Interview Scorecard — [Role Title] ([Level])

**Instructions:** Rate each criterion 1–4 using the behavioral anchors.

| Criterion | Weight | 1 — Not demonstrated | 2 — Below bar | 3 — Meets bar | 4 — Exceeds |
|-----------|--------|----------------------|---------------|---------------|-------------|
| [Must-have 1 — exact JD phrasing] | High | No evidence observed | Partial or inconsistent | Meets stated JD requirement | Significantly beyond requirements |
| [Must-have 2 — exact JD phrasing] | High | No evidence observed | Partial or inconsistent | Meets stated JD requirement | Significantly beyond requirements |
| [Nice-to-have 1] | Medium | No evidence | Mentioned, not demonstrated | Demonstrated | Deep expertise shown |

**Overall Recommendation:**
☐ Strong Hire  ☐ Hire  ☐ No Hire  ☐ Strong No Hire

**Notes / Observations:**
[Free text]

Field rules:

• Criterion names in the scorecard match JD must-have requirement phrasing exactly
• Each behavioral anchor is one sentence or phrase
• "Meets bar" (score 3) aligns directly with the stated JD requirement language
• Nice-to-have criteria row omitted if no nice-to-haves were provided

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: KirKruglov
• Source: KirKruglov/claude-skills-kit
• License: MIT

Install and usage instructions live in the source repository linked above.