feat: delete legacy page-intent audit

[cwjwisse]

Summary

• Delete src/page-intent/ directory (handler.js, prompts.js, sql/).
• Remove page-intent import + HANDLERS entry from src/index.js.
• Delete tests + fixtures.
• llmo-referral-traffic reader path (site.getPageIntents()) byte-unchanged.
• Coverage stays at 100% lines/branches/statements.

Context

Page-intent classification migrated to Mysticat (Blackboard producer + Projector). Audit hasn't run automatically since early March 2026.

Coordinated PRs

• adobe/mysticat-projector-service feat/blackboard-entity-fact-value (W5)
• adobe/mysticat-projector-service feat/page-intent-projector-config (W2)
• adobe/spacecat-shared feat/page-intent-drop-topic (W3)
• experience-platform/mystique feat/page-intent-enum-reconciliation (W1)
• Pending: adobe/mysticat-data-service DB migration drop page_intents.topic

Landing order

1. Mysticat pipeline (W1+W5+W2 + DB topic drop) deploys and produces rows in stage
2. Confirm llmo-referral-traffic Excel report populated for a paid test site
3. This PR merges (audit-worker handler removed)
4. Configuration PATCH: remove orphaned page-intent job + handler entries (ops)

Test plan

• npm test 8384 passing / 0 failing / 16 pending
• 100% coverage gate
• llmo-referral-traffic tests unchanged + passing
• Stage E2E: confirm page_intent column populates from Mysticat post-deploy

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[MysticatBot]

Hey @cwjwisse,

Strengths

• Exemplary change discipline. This is a pure deletion with zero functional additions or modifications mixed in. The scope is exactly one concern: remove the dead handler. The PR description documents the migration context, coordinated PRs, and landing order - this is how you decommission code in a distributed system.
• Complete removal from the registration path. Both the import statement (src/index.js:77) and the HANDLERS map entry (src/index.js:190) are removed, plus the package.json SQL directory reference. No dangling wiring left behind.
• Attack surface reduction. Removing the page-intent handler eliminates an LLM prompt injection surface (user-controlled page content passed to Azure OpenAI), an Athena SQL template interpolation surface, and S3 object read operations. All three trust boundaries are cleanly removed.
• Graceful degradation already handled. The SQS dispatcher returns a 404 with a log warning for unknown audit types. If any stale page-intent messages arrive after merge, they will be handled safely without crashes or retries.
• Downstream reader path unaffected. The llmo-referral-traffic handler's call to site.getPageIntents() is a database read via the data-access layer - completely independent of the deleted handler code.
• Tests removed with the code. The full 639-line test suite is deleted alongside the source. No orphan tests remain.
• No orphaned credentials or security controls. The deleted code consumed shared clients and env vars used by other handlers. No secrets, IAM permissions, or validation logic becomes orphaned.

Assessment

Ready to merge? Yes
Reasoning: This is a clean removal of dead code that was superseded by the Mysticat Blackboard/Projector pipeline. CI passes, coverage is unaffected (100%), the dispatcher handles unknown types gracefully, and the landing order ensures the replacement system is validated before this code is removed. The only remaining sequencing dependency (configuration PATCH to remove orphaned job entries) is documented ops work post-merge.

---

_{Skill: pr-review | Model: us.anthropic.claude-opus-4-6-v1[1m] | Duration: 0m 28s | Cost: $3.11 | Commit: d6c2c50a5f2a8b22a1b921c7fa2ab6ca624d6de3
If this code review was useful, please react with 👍. Otherwise, react with 👎.}