# Portal Path

> Check an external developer portal's navigation and its path to the first API call against the skill's portal
> design, then redesign the navigation and plan the path. Paste the navigation tree (one page per line, indented,
> with [login], [partner] and [ticket] markers) and the onboarding steps from landing to the first successful call
> (step | minutes); the browser classifies every page, checks that pricing and use cases read before login, holds the
> path to a time-to-first-call target set by the API's complexity class, places every self-service surface, ranks the
> friction fixes and build vs buy, and groups the gaps into a Priority 1/2/3 roadmap. Built on the SkillSafe skill
> @samber/developer-portal-design, from samber/developer-platform-skills (MIT), by Samuel Berthe.

https://portal-path.skillsafe.ai/

## What is free
- The classification: every page to one kind from a fixed enum (home, pricing, use cases, catalog, reference, guides, quickstart, keys, logs, sandbox, usage, billing, changelog, status, errors, support, search, assistant, team, SDKs, samples, community, roadmap, other), on the top, solution or bottom layer, for evaluators, engineers or both, with its access (open, behind login, partner-only, needs a ticket).
- The checks, over the six focus areas (Discovery, Decision-making, Onboarding, Go-live, Maintenance & troubleshooting, Community): pricing and use cases before login (gated is a sales-enablement failure), a solution layer, ungated docs, no signup before the docs, self-service keys without a ticket, a quickstart, a sandbox, a ready-to-run collection, request logs, usage, billing, team and roles, changelog, status, error catalog, support, search, an AI answer assistant, a named owner, one portal shape per auth model.
- Time to first call: a target from the class (under 15 minutes for simple REST with token auth, under 30 for moderate, under an hour for OAuth plus webhooks; Twilio's 5 minutes as context only), the path's minutes and the measured median against it, and the friction steps (verification, approval, sales, ticket, card, copying a key, an SDK install).
- The ladders: the friction rungs in efficiency order (ungated exploration, ready-to-run collection, auto-provisioned key, personalized onboarding) re-ranked by the interview; build vs buy (managed, assembled, custom - custom never promoted without a named owner, internal-portal frameworks ruled out); the maturity slide (project, product, critical business infrastructure - the page's own rule; Pronovix's rubric is proprietary).
- Exports: the checked tree (Markdown), the checks (.csv), the onboarding steps (.csv) and everything (.json).
- An internal service catalog is out of scope: the page says so and the runs are blocked.

## The metered lanes
Two tasks on gpt-terra, metered per run, routed by `task`. Input string fields: `task` (ia or path), `portal`, `tree`, `steps`, `carry` (path only: the redesigned navigation) and `facts`.
- `ia` replies with `summary`, `options` (`id` layered, split or docs_first, `name`, `tradeoff`, `recommended`), `tree` (`label`, `depth`, `layer`, `audience`, `gated`, `kind`, `note`), `placement` (`kind`, `where`, `entry`) and `caveats`.
- `path` replies with `summary`, `ttfc` (`target_minutes`, `rationale`), `steps` (`n`, `step`, `minutes`, `removes`), `fixes` (`rung`, `action`, `why`), `roadmap` (`priority`, `area`, `item`, `check`), `build` (`choice`, `why`) and `caveats`.
The page holds every reply to its own facts: exactly one recommended architecture and it is the page's leaning, pricing and use cases never gated, only the page's kinds, every surface placed in a page of the proposed tree; the page's target minutes (a moved target is put back), a path under it with the docs before sign-up, the fixes in the page's rung order, every fail check as a Priority 1 item, and the page's build choice.

## Links
- API tutorial: https://portal-path.skillsafe.ai/api.html
- Source skill: https://skillsafe.ai/skill/@samber/developer-portal-design
- Source repository: https://github.com/samber/developer-platform-skills