Summary of work done

• Acknowledged the scope mismatch: the prior work was a live fix plus guardrail, not a complete task rewire.
• Rewrote /Users/samaguiar/Documents/Claude/Scheduled/infra-daily-pulse/SKILL.md to canonical body version v2.0.
• Updated /Users/samaguiar/Documents/Projects/admin/scheduled-tasks/registry.json so the registered task description matches the new behavior.
• Archived the stale duplicate /Users/samaguiar/Documents/Claude/Scheduled/infra-daily-pulse/SKILL 2.md to _archive/SKILL-2-pre-v2-rewire-2026-05-23.md instead of deleting it.

What changed in the task

• The task is now origin-first: Cloudways origin logs, Cloudways services, WordPress writable paths, and CallRail/Salesforce attribution health are primary.
• Cloudflare and Sentry are now optional enrichments only.
• OAuth gaps must be reported as enrichment skipped, not degraded attribution, when origin and attribution checks completed.
• Every run must attribute 5xx bursts by status, path, IP count, and repeated PHP/proxy error signature.
• Every run must check /wp-json/sal/v1/callrail-webhook and [SAL ERROR] patterns because they affect attribution.
• Every run must check WordPress can read SAL/Salesforce config keys without printing secrets.
• Every run may perform a no-write Salesforce token smoke test from WordPress context if helper functions exist.
• Every run must check writable paths that have caused failures, especially wp-content/uploads/wpcode/cache.
• Normal scheduled mode remains read-only: it writes repair packets for live fixes unless manually invoked with an explicit repair instruction.

QA performed

• Validated registry.json with python3 -m json.tool.
• Confirmed the registry description now says origin-first infra and attribution pulse.
• Confirmed the skill contains v2.0, the required attribution-health block, origin-log attribution rules, and optional-enrichment language.
• Confirmed the duplicate stale skill file is archived, not active.