[wab]

Closes #94

Renames enjeu to challenge (posts#110): content type challenge, files in challenges/<lang>/. "Enjeu" stays the French UI word.

What

• scripts/schemas/stage-ids.js: the seven Stage ids in bow tie order, copied from the website (website#220). Frozen.
• scripts/schemas/challenge.schema.js: strict Zod schema for a challenge. Exports CHALLENGE_KEYS, challengeSchema, validateChallenge. It covers every section, cap, count, key, order and step from spec Spec: challenge content type and Team member expert fields #91. Cap errors read 263 characters, cap 240 (23 over).
• scripts/validate-repository.js: validates challenges/fr/*.md and challenges/en/*.md. A challenge must have an empty markdown body (frontmatterOnly), since every section lives in frontmatter.
• scripts/__tests__/fixtures/challenges/: six golden fixtures, one per challenge, built from the six prototype mockups.
• Tests: table tests on the schema (caps, counts, Stage ids, strictness, optional and required fields) and validateRepository on a fixture repository.

Decisions the spec left open

• Slugs are English, from the name, the same in both languages. The slug is the kebab-case of the challenge name: sales-experience, customer-experience, revenue-performance, sales-compensation, revenue-systems, revenue-ai. Cross-references keep using key; the slug only drives the URL. A golden test checks the fixture names; validateRepository does not enforce it yet (follow-up below).
• Two caps exceed spec Spec: challenge content type and Team member expert fields #91. case.headline is 120, the spec says 90: the Revenue AI headline is 98 characters, so the spec's own rule (longest mockup value plus about 20%) applies. method.tools allows up to 9, the spec says 0 to 6: Sales Experience shows 9 tools and Revenue Systems 7. The website layout (website#212) must allow for both.
• Body error path is the root ([]), printed as root by the CLI.
• Slugs (method.tools, case.story) carry no cap. method.tools is required and may be empty.
• Fixtures live under scripts/__tests__/fixtures/challenges/, not challenges/, so pnpm validate stays green until real content is written.
• Order follows the mockup navigation: se 1, cx 2, rp 3, sc 4, sy 5, ai 6.
• A workstream covering two method steps records the first step number.
• Bracketed placeholders from the mockups are kept verbatim in the fixtures.
• Revenue AI case eyebrow ("Nous l'avons fait chez nous d'abord") has no schema field. The website can key it on ai.
• Not yet allowed: partner (Agent window demo in the challenge schema #96) and aiBridge.demo (Partner block in the challenge schema #97). The strict schema rejects them until those tickets add them.
• Tool slugs without a tool file: fixtures cite claude, claap, chargebee, oneflow and full-enrich, none of which exists in tools/ yet. Challenge cross-references and uniqueness #98 (cross-references) has to create them or adjust the fixtures.

Questions for review (from spec #91)

• Do the French caps also apply to English text, or does English need its own measure?
• A5: case.story stays optional until Ocobo confirms a Story for the Revenue AI client case.

Left as follow-ups

• Enforce slug = kebab-case of name in validateRepository. It needs a decision on accents and apostrophes in future names.
• A YAML number in case.metrics[].value (value: 30) gives a type error, not a hint to quote it.
• Lengths count UTF-16 code units, so an emoji counts as two characters.

Tests

pnpm test: 349 passed. pnpm validate: all content valid.

[github-actions]

🔍 Preview: https://website-9sytew4bq-ocobo-22231b32.vercel.app

Built from website main with content from this PR's branch.