Capability Heatmap — People × Competencies

Engineer	Prompt Decomposition	Constraint Definition	Verification Discipline	Context Isolation	Review Judgment	Test-First Thinking	Bounded Agent Design
Elena V.	4 Str	3 Pra	2 Dev	3 Pra	2 Dev	4 Str	1 Eme
Marcus T.	3 Pra	3 Pra	3 Pra	2 Dev	3 Pra	3 Pra	2 Dev
Anya K.	2 Dev	2 Dev	1 Eme	2 Dev	1 Eme	2 Dev	1 Eme
Jordan P.	4 Str	4 Str	3 Pra	4 Str	3 Pra	4 Str	2 Dev
Sam R.	2 Dev	1 Eme	2 Dev	2 Dev	2 Dev	2 Dev	1 Eme
Priya D.	3 Pra	3 Pra	2 Dev	3 Pra	2 Dev	3 Pra	1 Eme

4 Strong 3 Practitioner 2 Developing 1 Emerging

Top 3 Coaching Priorities

01

Verification Discipline

Critical

Four of six engineers are at Developing or below. The team accepts plausible AI output without testing edge cases. This is the single highest-leverage improvement.

→ Run a 2-week "Prove It" sprint: no AI output is merged without at least one negative test case.

02

Bounded Agent Design

High

The entire team is Emerging or Developing. No one defines checkpoints in multi-step AI workflows. Cascading errors are inevitable as AI usage scales.

→ Workshop: design a 4-step agent pipeline with verification gates. Pair Sam and Anya with Jordan for a guided exercise.

03

Review Judgment

Medium

Half the team catches syntax but not reasoning quality. AI-generated architectural decisions pass review without challenge.

→ Introduce "reasoning review" prompts: for any function over 20 lines, reviewers must write a one-sentence design justification.

Mentor Map — Recommended Pairings

JP

SR

Jordan P. → Sam R.

Constraint Definition & Verification Discipline

Jordan is Strong in both areas where Sam needs the most support. They work on adjacent features, making pairing natural.

EV

AK

Elena V. → Anya K.

Prompt Decomposition & Test-First Thinking

Elena excels at decomposition and TDD. Anya will benefit from seeing how Elena structures prompts before coding.

JP

PD

Jordan P. → Priya D.

Context Isolation & Review Judgment

Priya is close to Strong in several areas — Jordan can push her over the threshold with targeted code review pairing.

Practice Plan, Next 6 Weeks

Week 1–2

2-hour workshop + 1-week practice challenge

Verification Bootcamp

Hands-on session focused on treating AI output as hypothesis. Every participant writes negative test cases for real AI-generated code from the past sprint.

Week 3–4

90-min design session + peer review

Bounded Agent Design Lab

Design multi-step AI workflows with explicit verification gates. Teams of two design, present, and critique each other's pipelines.

Week 5–6

60-min live review + ongoing 1:1 integration

Review Judgment Clinic

Live code review of AI-generated PRs. Focus on reasoning quality, not syntax. Practice writing design justifications for AI choices.

Manager Talking Points — 1:1 Guide

Elena V.

→ Acknowledge strong decomposition — ask her to mentor Anya on this
→ Discuss verification gap: "What would help you catch edge cases earlier?"
→ Propose stretch goal: design an agent pipeline with checkpoints

Marcus T.

→ Well-rounded but no Strong areas — ask what he wants to specialize in
→ Suggest he lead the review judgment clinic prep
→ Check in on workload — even skill distribution may indicate playing it safe

Anya K.

→ Be supportive — she has the most room to grow and may feel behind
→ Pair with Elena for decomposition, pair with Jordan for verification
→ Set small, weekly goals: one capability focus per sprint

Jordan P.

→ Highest performer — validate his work and ask what challenges him
→ Formally ask him to mentor Sam and Priya
→ Discuss path to architect: bounded agent design is his growth edge

Sam R.

→ Needs the most support — avoid overwhelming, focus on one capability at a time
→ Start with constraint definition: concrete, actionable, and immediately useful
→ Pair with Jordan twice per week for the first month

Priya D.

→ Close to breaking into Strong in several areas — acknowledge momentum
→ Focus on context isolation: she's closest to leveling up there
→ Ask her to present at the review judgment clinic

Squad Alpha