# Assess Domain Capability

> Use when the user wants a source-backed personal self-assessment or learning diagnostic for a field, including competency maps, questionnaires, local answer capture, scoring, charts, and job-readiness reports.

- Skill: `advancedcat/assess-domain-capability` (Agent Skill, multi-file: 6 files)
- Install (CLI): `npx skillmds@latest add advancedcat/assess-domain-capability`
- Raw SKILL.md: https://api.skillmd.com/api/skills/advancedcat/assess-domain-capability/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: AdvancedCat (https://skillmd.com/u/advancedcat)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/advancedcat/assess-domain-capability

---


# Assess Domain Capability

## Overview

Build a defensible domain capability diagnostic from current sources: research the field, confirm scope, create a competency model, generate a 35-45 item questionnaire, collect answers locally, score them, visualize results, and write a job-readiness report.

## Invocation

Use examples:

- `$assess-domain-capability AI Agent`
- `$assess-domain-capability 前端工程师岗位胜任力`
- `$assess-domain-capability create a job-readiness self-assessment for data analytics`

## Required Gates

1. **Scope confirmation:** If the field is broad, propose 2-3 narrower scopes and stop for user confirmation before generating the assessment.
2. **Live research:** Search current, source-backed material before defining competencies or questions. Prefer official standards, certification blueprints, vendor docs, reputable reports, curricula, and real job descriptions. Cite sources and access dates.
3. **Model confirmation:** Present the competency model and assessment mix before building the webpage if the user has not already approved them.
4. **Local persistence check:** Do not claim a browser-only page can write arbitrary local files. Use an existing backend, create a small local backend, or provide JSON export/download as a fallback.
5. **Evidence-first reporting:** Treat self-rating as calibration evidence, not the main score. Job-readiness conclusions must rely primarily on objective, scenario, and rubric-scored evidence.

## Workflow

1. Collect intake: target role, target level, region/industry, use case, language, answer time, and whether the report should be strict or lightweight.
2. Load `references/assessment-design.md`, then perform live research. Keep concise research notes with source ids.
3. Build 4-8 competency dimensions with 3-6 subcompetencies each. Include behavior anchors from entry to expert-candidate levels.
4. Generate 35-45 items. Use self-rating only for confidence, familiarity, or frequency; use scenario, knowledge, and subjective case items for the main evidence.
5. Load `references/artifact-schema.md` when creating assessment JSON, answer storage, a local web app, or charts.
6. Build the webpage in the current project style. If product-design skills/tools are available and the UI brief is under-specified, use `product-design:get-context` first. Keep the first screen as the usable questionnaire, not a landing page.
7. Save answers locally using the schema reference. For subjective answers, score against the rubric before final scoring or mark them pending.
8. Run `scripts/score_assessment.py` against the assessment and response JSON. Use the JSON output for radar/bar charts and the Markdown output as a report scaffold.
9. Write the final report: strengths, weak areas, calibration gap, role-readiness label, evidence confidence, and 2-week/1-month/3-month improvement plan.

## Scoring Command

```bash
python3 skills/assess-domain-capability/scripts/score_assessment.py \
  --assessment data/assessments/<slug>/assessment.json \
  --responses data/responses/<slug>/<timestamp>-response.json \
  --output data/reports/<slug>/<timestamp>-scores.json \
  --markdown data/reports/<slug>/<timestamp>-report.md
```

## Common Mistakes

- Generating questions before research.
- Letting 1-10 self-ratings dominate the result.
- Asking broad, socially desirable questions instead of scenario choices.
- Producing a polished UI that cannot save answers locally.
- Reporting "you are senior" without scope, confidence, evidence, and caveats.