chore(pipeline): add project-local adversarial coding pipeline (Stage A/Unity)

[mlsamuelson]

Self-contained .kiro pipeline seeded from ddt-agent-skills-library-foundry: 5 acp-* agents (orchestrator/architect/coder/reviewer/visual), foundry + ASU skills (jira-figma-intake, unity-components, asu-design-a11y, asu-brand, unity-handoff-out, visual-diff), project-local MCP (Atlassian RO, Figma, Storybook), steering, hooks, preflight + handoff validator, README. Agents stop at branch+commits; humans open PRs.

TEST

• Create a new branch based on this.
• Ensure you have all the "One time setup" steps documented in the PR completed.
• Create a Jira ticket with a link to a single Figma design OR take a screenshot of the design and add it to .intake/
• Run Kiro: kiro-cli --agent acp-orchestrator
• Paste this as your first message to the orchestrator:

  Component: [REPLACE WITH SHORT COMPONENT NAME]. Build/validate a Unity [COMPONENT NAME] component that
  [REPLACE WITH SHORT DESCRIPTION].
  Intake: no Jira, no Figma — use the screenshot in .intake/HeroWithExpandableHeroes.png. [OR GIVE IT THE JIRA/FIGMA INFO]
  Target: pick the right Unity package and tell me your reasoning before coding.

What to expect:

1. INTAKE — it reads the screenshot, writes .pipeline/intake-brief.md, summarizes, and asks clarifying questions (e.g. new component vs. extend an existing component; which package). Answer them.
2. PLAN [your gate] — architect writes .pipeline/design-doc.md citing unity-components/asu-brand/asu-design-a11y. Review and reply "approved" to proceed (the only stop).
3. IMPLEMENT → REVIEW → VISUAL — coder (TDD, tokens, HTML-parity) → reviewer (3-phase) → acp-visual screenshots the render vs. your screenshot. Up to 3 fix cycles, autonomous.
4. GATE + HANDOFF — build+lint, run-story-tests, HTML-parity, VISUAL_PASS; writes ../handoff// and commits to feature branch.

[asu-jenkins-devops]

Storybook deployed at https://unity-uds-staging.s3.us-west-2.amazonaws.com/pr-1732/index.html

[mlsamuelson]

Overall, the full pipeline development, first pass is completed:

  All 14 tasks complete. Here's the wrap-up.

  What was delivered

  A repeatable, two-stage adversarial coding pipeline, validated end-to-end on a real component (Expandable Heroes):

  - Stage A — asu-unity-stack/.kiro/: foundry-seeded acp-orchestrator/architect/coder/reviewer/visual (model-diverse), ASU skills (jira-figma-intake, unity-components, asu-design-a11y, asu-brand, unity-handoff-out, visual-diff), project-local
  MCP (Atlassian RO, Figma, 8 Storybook), preflight + handoff validator, steering, README. Proven: flexible screenshot intake → design gate → TDD → GLM reviewer → vision visual-diff → 5 cycles → PR-ready gate → valid handoff.
  - Handoff (chain/handoff/<anchor>/): handoff.json + markup.html, schema-validated, peer to both repos.
  - Stage B — asufactory1/.kiro/: same agents + webspark-handoff-in, unity-consume, webspark-drupal + synced design/brand skills, MCP (Atlassian RO, GitLab RO, Figma). Proven: consumed the handoff → renovation-theme block + profile config +
  hook_update_N + Kernel/Playwright tests, Drupal core untouched.
  - Docs: chain/pipeline-plan.md (the approved plan) and chain/pipeline-runbook.md (operations + all conventions/findings).

  Your corrections this turn are encoded: Drupal-core (not Webspark-core) rule, the three placement patterns, the websparkci.asu.ddev.site test site, @websparkci.local drush rule, and the verification gating (Playwright-blocked ⇒ BLOCKED;
  visibility-pair bounding-box assertions; DELIVER asset-fingerprint curl ⇒ P1 DEPLOY_PENDING).

  Open operational actions (yours; agents don't push/PR)

  - Unity: git push --force-with-lease origin feat/expandable-heroes (clean component PR), then open the PR; push pipeline-trial.
  - Webspark: open the MR from feat/expandable-heroes; push pipeline-trial.
  - Ensure JIRA_USERNAME/CONFLUENCE_USERNAME are exported (replacing the old JIRA_USER).
  - Optional later: upstream the asu-brand skill (and a unity-webspark pipeline) to the foundry.

[mlsamuelson]

Kiro developed code. I've reviewed. See no concerns. Successful run evidenced in #1733

[davidornelas11]

Good! Any suggestions will be added on an as needed basis