flow

Product discovery · Portable skill

align/

Check whether the product still fits its goals and users and what the outcome evidence shows; read-only.

npx skills add gaboesquivel/flow-skills -s align

Say it

Natural language is the interface. Any of these, or your own words, reach the skill.

  • “check alignment”
  • “are we building the right thing”
  • “did it land”
  • “measure our progress”

When to use it

The product compared with its goals, problems, and metrics, and each outcome with the evidence ("did it land", "are we building the right thing"); shipped with impact unknown is a valid answer.

The skill

The authoritative, versioned instruction your agent follows, from skills/align/SKILL.md.

Are we building the right thing, and what does the evidence say? "Check alignment" compares what the product is with what it is meant to be, and each outcome with the evidence you can reach. This is the what and the why, never the how: code quality is review, shaping a new idea is explore, and updating documents is sync-docs.

Gather

A missing source is a finding, not a reason to stop or ask.

  • Intent: PRODUCT.md (problem, users, core journeys, first useful release, main and supporting metrics, kill criterion, riskiest assumptions, alternatives, non-goals), DESIGN.md for journey steps, accepted specs and decisions, and any roadmap or candidate stubs. A Proposed record or draft spec is planned, not intended fact. For "did it land", the shipped spec's Outcome line names the evidence to look at and the decision it informs.
  • Reality: code, routes, commands, and tests show what users can do; recent Git history shows where effort went. Never count a capability as implemented on a document's word alone.
  • Evidence: metric data through whatever the host reaches: tools, local data, or exports the user supplies. Nothing is required. Measure only what is defined, and say when a definition is missing.

Judge fit

Map each core journey and stated problem to the capabilities that serve it, each implemented, partial, planned, or none. Then:

  • Working: journeys the implemented product serves. A shipped capability is not demonstrated value until evidence shows it: "did it land" asks about the outcome of landed work, never to merge it, and separates delivered (ship's liveness) from used and from the outcome moved, and leaves what no source shows unknown. "Shipped, impact unknown" is a valid answer; say what evidence would settle it and which decision waits on it.
  • Missing: stated problems with nothing behind them, and metrics nothing in the product can observe.
  • Drift: built work no stated problem asks for, or docs describing a product the code no longer is.
  • Complexity: surfaces that cost more than the problem they serve.

Judge each metric

  • Result: baseline, current value, target, source, period, sample size, and limitations, with the calculation shown. Count each qualifying outcome once; an emitted event is not a successful outcome unless the definition says so.
  • Compare only compatible definitions and populations; a before/after difference alone is not causation. Say when evidence is a fixture, sample, or test data.
  • Gap, when there is no conclusion, blocking one first: not collected; instrumentation missing or wrong (nothing computes it, or collection misses, double-counts, or misattributes, such as human and agent traffic); no access; or insufficient (too small, short, or noisy).

Unknown is never zero. Missing access never hides what can be established; never invent a value. For each gap, name the smallest response: reuse existing collection, then a manual or export count, before any new service.

Report

Lead with a visual: the journey-to-capability map, and a metric table (metric, result, target, source, gap) when there is more than one. Then working, missing, drift, and complexity in a few lines each, citing evidence and separating fact from inference. End with up to three priorities ranked by impact on the stated vision and metrics. Omit empty sections; say when intent is too thin to judge.

Boundaries

Read-only: never edit, install tools, connect accounts, change tracking, or schedule work. Name the skill that would act (sync-docs, explore, to-spec, build) and wait.

Jump to, or try a question