assess — whole-app health

← Book index · a five-expert panel over your product

What it is: the app-level review — not a diff, not a screen, the product: how features link, which are strong, which are thin, and where you stand against the market.

How it works

flowchart TD
  E[evidence base: code · feature changelog ·<br/>measured usage · docs] --> P{{five expert lenses}}
  P --> L1[domain expert:<br/>real workflows served?]
  P --> L2[software engineer:<br/>coupling · fragile areas]
  P --> L3[PM: incomplete journeys ·<br/>features that don't compose]
  P --> L4[market analyst:<br/>cited, today's date]
  P --> L5[improvement lead:<br/>value vs effort synthesis]
  L1 & L2 & L3 & L4 & L5 --> R[assessment.md:<br/>feature map · Strong · Needs improvement ·<br/>market position · next moves]
  R --> H[your pick → define/blueprint ⛔]

Best cases

• "Which features are weak?" — periodic product health checks.
• Before planning a big quarter — its ranked "next moves" feed discover/define.
• Inherited products — a new owner's honest map of what's actually there.

Example

itqan:assess "analyze my app — feature linkage, strong vs weak, vs the market"

What you get

assessment.md: a feature-linkage map · Strong (with evidence — it says what's good, not just what's wrong) · Needs improvement ranked by value-vs-effort with the concrete fix and which expert flagged it · a cited market position. Optionally a throwaway prototype of the top recommendation so you can feel it before deciding.

Hand-offs

Approved improvements → define → blueprint, stopping at the normal plan gate. Net-new ideation leads with discover; code-level review with inspect; security depth with harden. Reuses the run's multi/single-agent answer instead of re-asking.

Pro tip: it asks two things up front — panel as parallel agents or one, and report-only vs report+prototype. Pick report-only the first time; add the prototype when a finding is worth feeling.