feat(persona): the demand side — sourced reader model + challenge gate (v0.5.0)

[GabrielAmz]

What kind of contribution?

• New skill
• Update to existing skill
• Docs / README

If new skill, the 5-step checklist

• Workflow-skill naming (persona, noun-style like moodboard; the wa- prefix predates the v0.2 rename)
• Directory name exactly matches the name: field in SKILL.md front matter
• description: is trigger-phrase dense, ends with cross-references to related skills
• Body is under 500 lines — research + challenge protocol detail pushed into references/persona-method.md
• Loads .agents/webanatomy-context.md at Step 1 (offers webanatomy-setup, never blocks)
• Reference patterns included (benchmark FAQ gallery as objection inventories; G2/Capterra/Trustpilot as review-mining sources)

What does this skill do?

Adds the demand side to a pack that only modeled the supply side. Build mode ("build a persona", "what are my customers' objections") researches a sourced reader model into .agents/webanatomy-persona.md: ranked objections with source tags (user-stated > review-mined > inferred, never invented), the reader's vocabulary, proof preferences, optional buying committee, and a competitor claims map. Challenge mode ("cold read this as my customer", "would my customer buy this") reads any live page, write-page draft, or build-page wireframe as that reader — five-second read, scroll narrative, objection ledger, credibility flags, click decision — and writes challenge.json, a pass/fail handoff write-page consumes to rewrite only what failed (max two loops).

Wiring into the existing pipeline (no parallel pipeline):

• audit-page 0.4.0 — Step 4.5 objection coverage when the persona file exists; optional objectionCoverage block in audit.json (schema stays v2).
• write-page 0.6.0 — copy angles labeled with the objection they answer, persona vocabulary feeds the customer-language rule, claims-parity check, FAIL challenge as highest-precedence brief in Step 1.5.
• webanatomy-setup 0.4.1 — suggests persona as a next move.
• Pack bumped to 0.5.0 via ./bump-version.sh; VERSIONS.md entry added, roadmap renumbered; orchestration doc gains Part 3 recording why this is one sourced persona and not a five-judge copy tournament (the information test).

After merge: git tag v0.5.0 && git push --tags && gh release create v0.5.0 --target main.

Test plan

• Ran ./validate-skills.sh — 8/8 pass
• Manually invoked from Claude Code on a real project — pending a live run (build mode on Web Anatomy itself is the natural first test)
• Cross-references to related skills are valid (persona ↔ setup/audit-page/write-page/build-page/find-examples all exist)

🤖 Generated with Claude Code