talent-assess

/talent-assess

Run a Talent-Augmenting Layer onboarding assessment. Profile is always saved locally. MCP tools used for scoring when available.

Flow

1. Read the assessment protocol from mcp-server/src/assessment.py (question banks SECTION_A_QUESTIONS, SECTION_B_QUESTIONS, SECTION_D_QUESTIONS, ESA_ANCHORS, domain taxonomy via suggest_domains()).
2. If a profile already exists in profiles/pro-*.md or profiles/tal-*.md for this user, note it and confirm they want to replace it.
3. Run the assessment conversationally, one section at a time.
4. Compute scores:
   • If MCP tools are available, call talent_assess_score with the raw answers.
   • Otherwise, use the formulas in assessment.py directly.
5. Generate the profile markdown (use generate_profile_markdown() template from assessment.py).
6. Save to profiles/pro-{name}.md (lowercase, kebab-case).
7. Present scores and ask: "Do these feel accurate? Anything you'd adjust?"

Assessment sections

• Identity: name, role, organisation, industry, context summary
• Section A (AI Dependency Risk): 5 questions, behavioural anchors 1-5
• Section B (Growth Potential): 5 questions, behavioural anchors 1-5
• Section D (AI Literacy): 4 questions, behavioural anchors 1-5
• Expertise Domains: suggest domains via talent_suggest_domains (or the taxonomy in assessment.py), rate each 1-5 on the ESA scale, validate with behavioural examples. Aim for 5-10 domains.
• Goals & Context: career goals, skills to develop/protect, task classification (automate/augment/coach/protect/hands-off), red lines, learning/feedback/communication style

Guidelines

• Be warm and conversational, not clinical
• Explain WHY you're asking each question briefly
• If the user gives a vague answer, probe with the behavioural anchors
• It's OK to adjust scores based on behavioural evidence (expert interviewer mode)
• The whole assessment should take 15-20 minutes
• If a profile already exists, note it and confirm they want to replace it