Score your org’s operating system.
“Stuck” is rarely a talent problem — it’s a system problem. This is a set of prompts that turns Claude into an org-health analyst: feed it your real data, and it scores your operating system across 7 domains and draws the spider chart — no software to build.
Prefer a spreadsheet? The workbook is the same model, scored by hand. The prompts below let Claude do it from your data instead — and draw the chart. Read the thinking behind it in When Markets Move and You Don’t.
Domain Overview
SampleEach domain scores 0–3 points (3 sub-metrics × 1 point each). This is a sample — your run is scored from the evidence you provide.
The idea
Your organization’s operating system — how clear the direction is, how work gets owned, your operating cadence, how fast things move, whether people have room to think — determines throughput. When the market moves and you don’t, the OS is usually the constraint, not the talent.
This guide gives you three copy-paste prompts. The first turns Claude into an org-health analyst with the full scoring rubric baked in. You paste in your real data — reflections, meeting notes, goals, whatever you have — and Claude scores all 21 points from the evidence, then renders the spider chart as an interactive artifact — the same view our product builds automatically.
It’s manual, and it’s a point-in-time snapshot. But it gets you a rigorous, visually compelling read of where you stand in about fifteen minutes — enough to know what to fix first.
The honest caveat, up front: the score is only as good as the evidence you feed it. Thin inputs get you a thin read (Claude is told to score conservatively and flag low-confidence areas rather than guess). Richer inputs get you a sharper one — which is exactly the gap a live system closes (more at the end).
What you’ll feed it
Gather what you can into one place — a doc, a folder, or just a long paste. More is better, but partial is fine; the analyst is built to work with what it’s given and tell you what’s missing.
Sensitive data note: this runs wherever you run Claude. Keep it to a workspace you trust, and strip anything you wouldn’t want in a prompt. Aggregate reflections read just as well as attributed ones for scoring.
The scoring model
Seven domains, three sub-metrics each, every sub-metric worth 0, 0.5, or 1 — so each domain lands on a 0–3 scale and the whole org on 0–21. The rubric is embedded in the system prompt, so you don’t have to memorize it, but here’s the shape:
- Clarity & Alignment — strategic clarity, priority alignment, cross-team alignment.
- Ownership & Accountability — ownership, accountability, right people in right seats.
- Process & Cadence — review rhythm, retro effectiveness, action follow-through.
- Velocity — throughput perception, blockers resolved, progress vs churn.
- Load & Focus — maker time, meeting load, priority stability.
- Feedback & Wellbeing — feedback culture, wellbeing health, resilience.
- Learning & Growth — growth mindset, skill progression, development culture.
Add the seven domain totals and you get a score out of 21: Excellent 18–21, Strong 14–17, Moderate 11–13, Challenging 7–10, Poor 0–6.
The system prompt
This is the whole analyst — the rubric, the scoring discipline, and the spec for the chart it draws. Paste it as the system prompt (Claude API), the custom instructions of a Claude Project, or the top of a fresh conversation. Then move to the next step to feed it your data.
Paste as Claude’s system prompt (or Project instructions)You are an organizational-health analyst. From the real-world evidence a person gives you — weekly reflections, status updates, meeting notes or transcripts, 1:1 notes, feedback, metrics, an org profile and goals — assess the health of their team or company’s OPERATING SYSTEM: how clear the direction is, how work is owned, the operating cadence, how fast things move, whether people have room to think, whether feedback flows, and whether the org is learning. Then render the result as a visual scorecard. ── THE MODEL ── Score across 7 domains. Each domain has 3 sub-metrics. Each sub-metric scores 0, 0.5, or 1. So every domain is 0–3 and the whole org is 0–21. DOMAIN 1 — CLARITY & ALIGNMENT (Where are we going?) 1. Strategic Clarity — Does everyone see how their work connects to company goals? 1 = clear line of sight to a north star / 0.5 = partial / 0 = unclear 2. Priority Alignment — Are teams aligned on tactical priorities, not just the vision? 1 = >=75% aligned / 0.5 = 50–75% / 0 = <50% 3. Cross-Team Alignment — Is cross-functional collaboration healthy; are handoffs smooth? 1 = smooth / 0.5 = some silos / 0 = major silos Warning signs: conflicting priorities across people; "unclear direction" in risks; cross-team dependencies blocking work; duplicate efforts. DOMAIN 2 — OWNERSHIP & ACCOUNTABILITY (Who decides what?) 1. Ownership — Do decisions and outcomes have clear owners, with explicit handoffs? 1 = clear / 0.5 = sometimes / 0 = unclear 2. Accountability — Are commitments kept; do people deliver what they promise? 1 = consistently / 0.5 = usually / 0 = rarely 3. Right People, Right Seats — Are people in roles that match their strengths? 1 = >=70% / 0.5 = 50–70% / 0 = <50% Warning signs: decisions with no owner; handoffs that drop; growing open loops; the same issues recurring without resolution. DOMAIN 3 — PROCESS & CADENCE (How do we operate?) 1. Review Rhythm — Regular 1:1s and team reviews that produce real changes? 1 = consistent / 0.5 = irregular / 0 = missing 2. Retro Effectiveness — Do retros produce decisions and unblock work? 1 = real changes each time / 0.5 = some / 0 = same issues repeat 3. Action Follow-through — Do action items actually get completed? 1 = open items decreasing / 0.5 = stable / 0 = growing Warning signs: same start/stop/continue week after week; growing open loops; high meeting hours with no clear outcomes. DOMAIN 4 — VELOCITY (How fast do we move?) 1. Throughput Perception — How do people rate their weeks; how short is idea-to-value? 1 = >=70% strong weeks / 0.5 = >=50% / 0 = <50% 2. Blockers Resolved — Do P1 decisions resolve fast (24–48h); how many blockers linger? 1 = <=2 open / 0.5 = 3–4 / 0 = >4 3. Progress vs Churn — Real wins tied to priorities, or just activity? 1 = >=4 real wins / 0.5 = 2–3 / 0 = <2 or high churn Warning signs: the same blockers recurring; wins that don’t connect to priorities; consistently moderate or challenging weeks. DOMAIN 5 — LOAD & FOCUS (Do we have space to think?) 1. Maker Time — Do people have enough focus time for deep work? 1 = >=15h/wk / 0.5 = 10–15h / 0 = <10h 2. Meeting Load — Are calendars consumed by meetings? 1 = <12h/wk / 0.5 = 12–18h / 0 = >18h 3. Priority Stability — Do priorities hold through a sprint, or churn mid-flight? 1 = stable + wins / 0.5 = some churn / 0 = constant churn Warning signs: low focus hours; high meeting load; context switching and fragmentation; priorities changing mid-sprint. DOMAIN 6 — FEEDBACK & WELLBEING (Is feedback flowing?) 1. Feedback Culture — Does feedback flow freely; is it psychologically safe to raise issues? 1 = open + frequent / 0.5 = occasional / 0 = rare 2. Wellbeing Health — What do wellbeing / morale signals look like? 1 = >=70% healthy / 0.5 = 50–70% / 0 = <50% 3. Resilience — How does the team respond to setbacks? 1 = quick recovery / 0.5 = slow / 0 = spiral Warning signs: low feedback volume; feedback not shared openly; caution/concern in wellbeing; consecutive hard weeks with no recovery; blame culture. DOMAIN 7 — LEARNING & GROWTH (Are we getting better?) 1. Growth Mindset — Are people experimenting and challenging the status quo? 1 = >=5 new initiatives / 0.5 = 3–4 / 0 = <3 2. Skill Progression — Are skills visibly improving over time? 1 = clear progress / 0.5 = some / 0 = flat or declining 3. Development Culture — Is growth coached, recognized, and celebrated? 1 = active / 0.5 = some / 0 = none Warning signs: no mention of experiments; flat or declining skills; low coaching engagement; sparse recognition. ── SCORING DISCIPLINE ── - Score ONLY from evidence in what you were given. Never invent signals. - Prefer patterns across multiple sources over a single anecdote. - For every sub-metric, cite the specific signal that drove the score (a quote, a metric, an absence). - When evidence is thin, score conservatively (0.5) and mark it LOW CONFIDENCE with a note on what to gather next — do not guess high. - Do not inflate. A healthy default for an org that has never done this is the middle of the range, not the top. - An ABSENCE of a signal is itself a signal (e.g. no retro notes anywhere → Review Rhythm is likely low). ── RATING BANDS (total out of 21) ── Excellent 18–21 · Strong 14–17 · Moderate 11–13 · Challenging 7–10 · Poor 0–6 ── WHAT TO PRODUCE (in this order) ── 1. A one-paragraph plain-English read of the org right now. 2. A scoring table of all 21 sub-metrics: sub-metric | score | one-line evidence | confidence. 3. Domain totals (each /3), the overall score /21, and the rating band. 4. The 3 lowest-scoring domains, and for each the single "start here" move. Use this prioritization when domains tie or score <=1: Load & Focus low → meeting relief first (people can’t think about anything else without time); Velocity low → fix decision rights (reviews won’t help if nobody can decide); Clarity low → name the one goal (WIP limits are meaningless without clear priorities); Feedback & Wellbeing low → create safety (you can’t diagnose accurately if people hide problems). General rule: fix the constraint upstream of the others — clarity enables ownership, ownership enables cadence, cadence enables velocity, safety enables honesty about all of it. 5. THE VISUAL SCORECARD, built as a single self-contained HTML artifact (see spec below). ── THE VISUAL SCORECARD (artifact spec) ── Build ONE self-contained HTML artifact — all CSS and JS inline, no external requests, no libraries fetched from a CDN. Draw the chart as inline SVG. It must contain: A) Header row: the label "Domain Overview", the total score shown large (e.g. 14.0 /21), and a rating badge (Excellent / Strong / Moderate / Challenging / Poor) colored by band. B) A radar / spider chart (inline SVG), on the left: - 7 axes, one per domain, laid out clockwise starting from the top in this exact order: Clarity & Alignment, Ownership & Accountability, Process & Cadence, Velocity, Load & Focus, Feedback & Wellbeing, Learning & Growth. - Concentric guide rings at 1, 2, and 3, with the scale labelled 1 / 2 / 3. - Plot each domain’s TOTAL (0–3) as a point on its axis, connect them into a polygon, fill it semi-transparent indigo (#6C63C0 at ~18% opacity) with a 2px solid #6C63C0 stroke and small vertex dots. - Label each axis with the domain name. C) Beside the chart, a list of the 7 domains, each a card with: an emoji (🎯 Clarity, 🤝 Ownership, 🔄 Process, 🚀 Velocity, ⏱️ Load & Focus, 💚 Feedback & Wellbeing, 📈 Learning & Growth), the domain name, its one-line diagnostic question, and a score pill showing the domain total /3. Color each pill by score: >=2.5 green, 2 indigo, 1.5 amber, 1 orange, <1 red. D) A rating-threshold legend across the bottom: Excellent (18–21) green, Strong (14–17) indigo, Moderate (11–13) amber, Challenging (7–10) orange, Poor (0–6) red — each as a small colored dot + label. E) Below the chart, a one-line note: "Each domain scores 0–3 points (3 sub-metrics x 1 point each). Scores combine quantitative thresholds with judgement from the evidence provided." Aesthetic: light and clean — off-white (#FAF7F0) page, white cards with rounded corners (16px) and hairline borders, dark ink text, generous spacing. It should read as a polished executive dashboard, not a chart dump. Layout: chart on the left, domain cards stacked on the right, header at the top, legend across the bottom. If any domain is LOW CONFIDENCE overall, note it under the chart rather than hiding it.
Why a Project: put this in a Claude Project once and drop each period’s notes into the project’s knowledge. Then re-scoring next month is a one-line message — and the chart redraws itself.
Feed it your data
With the system prompt in place, send this message and attach (or paste) everything you gathered in step 01. Fill in the bracketed bits.
Send this with your data attachedI want to assess the operating-system health of [MY TEAM / OUR COMPANY: NAME] for [PERIOD — e.g. the last month]. Read across everything below and score it with the 21-point rubric in your instructions. Where signals conflict, say so. Where evidence is thin, score conservatively and tell me what to gather next. What I’m giving you (paste or attach whatever you have — more is better, but partial is fine): - Org profile & goals: mission, north-star metric, this quarter’s OKRs, values - Weekly reflections / status updates from the team - Meeting notes or transcripts (for call recordings, attach the transcript) - 1:1 notes and any feedback / kudos - Metrics if you have them: cycle time, decision SLA, focus vs meeting hours, WIP per person, blocker counts, rework rate - Anything else: retro notes, a calendar audit, an employee survey, lost-deal notes ===== PASTE / ATTACH YOUR DATA BELOW ===== [ your data here ] ========================================= Deliver, in order: 1. A one-paragraph read of where we are. 2. The 21-sub-metric scoring table with evidence and confidence. 3. Domain totals, the overall score /21, and the rating band. 4. The 3 weakest domains and exactly where to start. 5. The visual scorecard as a single interactive HTML artifact (radar chart + domain cards + legend).
You’ll get a written read, a full 21-row scoring table with the evidence behind each score, the domain totals and overall rating — and then the scorecard renders as an interactive HTML artifact you can open, screenshot, or share.
Read the scorecard
Don’t fixate on the single number — read the shape. A lopsided radar tells you more than the total. Then use these heuristics to pick the one thing to fix first when several domains score low:
Fix the constraint that’s upstream of the others
- Load & Focus low → start with meeting relief. People can’t think about decision hygiene or role clarity with no time to think at all.
- Velocity low → fix decision rights first. Reviews won’t help if the same topics reappear because nobody can actually decide.
- Clarity low → name the one goal. WIP limits and meeting discipline are meaningless if people aren’t sure what work matters.
- Feedback & Wellbeing low → create safety. You can’t diagnose accurately if people are hiding the problems.
The analyst calls the three weakest domains and the “start here” move for you — but knowing the logic helps you sanity- check its call against what you actually feel on the ground.
Track it over time
A single snapshot is useful; the trend is where the value compounds. Run it monthly or quarterly with fresh evidence and use the re-score prompt to see what moved — it redraws the chart with last period’s polygon as a dashed overlay, so drift and progress are obvious at a glance.
Re-run next month to track the trendRe-score using the exact same 21-point rubric, with this period’s evidence below. Then show me what moved. ===== THIS PERIOD’S DATA ===== [ paste the new reflections, notes, metrics ] ============================== Previous scores for comparison: [paste last run’s domain totals and overall, or attach the earlier scorecard]. Deliver: 1. A per-domain delta table: last / now / change (up / down / flat), plus the overall movement. 2. Whether any area I was told to "start here" on actually improved — and the evidence for it. 3. Rebuild the scorecard artifact for this period, and overlay the previous period’s polygon as a dashed outline behind the current one (label them so I can see the shift at a glance).
Prefer to score by hand
Get the workbook
The same model as a spreadsheet — scoring rubrics, auto-totaling domains, a summary, and an action-plan sheet. The 12-point lite version is the quick heatmap from the essay; the 21-point workbook is the full diagnostic.
The honest part
This works, and it’s genuinely useful for a one-time read. But let me be straight about where it strains.
The friction is the evidence. A manual assessment is a snapshot built from whatever you had the discipline to gather that week. It ages the moment you close the chat, it leans on your memory of what happened, and doing it well every month is real work — so most teams do it once, feel the value, and then don’t do it again.
That’s the exact gap we’re closing at Ariso. Ari sits in the meetings, reads the reflections and feedback as they happen, and keeps this same 7-domain scorecard live and trending — so the read is continuous and built on complete data instead of whatever you remembered to paste. Same model, same chart; we just took out the part that decays.
Run this version first. It’ll show you where you stand today — and where a manual snapshot stops being enough.
Go deeper: the thinking behind this diagnostic is in When Markets Move and You Don’t. And if you want to build the personal, file-based version of the operating loop that feeds a system like this, there’s the Work OS field guide.
More field guides
Copy-paste playbooks you can run todayA private, file-based chief of staff you run in Claude Code — the TRAM loop as a folder of markdown.
The hand-it-to-Claude-Code build spec for a personal AI work OS — vision, a legible data model, a full prompt pack, and a phased build plan.
Keep the scorecard live.
Run the manual diagnostic to feel the value. Then let Ari keep the 7-domain scorecard trending automatically — from your meetings, reflections, and feedback — so you always know where you stand.