← All runbooks
aapersh / diagnosis-and-framing★ Featured · worked examples

Diagnosis & Framing

Produce a defensible, fact-first strategic diagnosis that a leadership team or board can act on. Do the work in the order a top-tier consultant would: build the baseline before debating solutions, isolate the one or two constraints that actually bind, surface…

agent claude-codemodel claude-sonnet-4-6snapshot python312-uveval programmatic5 stepsv1.0.0

Deploy Diagnosis & Framing to your jetty.io

One-click installs this runbook into a collection on your Jetty account. You can run it from the Spot dashboard, schedule it, or pipe inputs in via the API.

Deploy on jetty.io →View source
Run time3–6 mins
Headline outputdiagnosis.md

Runs on Jetty's managed sandbox. No setup. Free for your first 10 runs.

Worked examples · 2

Real runs, real outputs.

Meridian — should we raise a $40M Series B to go upmarket?
diagnosegrowth-deceleration-diagnosiseval ✓

Meridian — should we raise a $40M Series B to go upmarket?

Structural (not cyclical) deceleration across win rate, CAC, NRR and churn while a PE-backed competitor bundles a free substitute. The diagnosis isolates the binding…

diagnosis.mdsummary.mdvalidation_report.json
claude-sonnet-4-6 · 6 minView run →Source
Lumen — pivot to B2B, or fix the consumer funnel?
diagnosepivot-vs-focus-decisioneval ✓

Lumen — pivot to B2B, or fix the consumer funnel?

A stalled consumer subscription app weighing two inbound enterprise LOIs against fixing its funnel. The diagnosis isolates 8.5% monthly churn as the single binding…

diagnosis.mdsummary.mdvalidation_report.json
claude-sonnet-4-6 · 4 minView run →Source
The shape of the run

5 steps · start to finish.

  1. 1
    Step 1

    Setup & Input Guard

    ▶
    mkdir -p {{results_dir}}
    • Read {{situation}}. If it is empty, whitespace, or still the literal placeholder, fail fast: write validation_report.json with stages[0] = {name:"setup", passed:false, message:"situation input missing"}, overall_passed:false, and stop.
    • Read {{decision}}. If empty, infer the single most consequential decision the situation implies and record it; you will state it explicitly in the Executive Read.
    • If init_params.file_paths is non-empty, read each file and fold its facts into the fact base (label them by source).

    Do not start recommending yet. The next step is to separate what you know from what you believe.

  2. 2
    Step 2

    Build the Fact Base (separate facts, interpretations, assumptions)

    ▶

    Before any analysis, sort everything the situation gives you into three labelled buckets:

  3. 3
    Step 3

    Write `diagnosis.md` (the three methods, composed)

    ▶

    Write {{results_dir}}/diagnosis.md with exactly this structure. Every section is required; tables must have the named columns (the Step 4 validator checks for them).

  4. 4
    Step 4

    Evaluate, Validate & Iterate (max 3 rounds)

    ▶

    Run a programmatic structural check on diagnosis.md and write the result. This does not grade prose taste — it enforces the method's non-negotiables so a weak pass can't slip through.

  5. 5
    Step 5

    Write Executive Summary

    ▶

    Write {{results_dir}}/summary.md: