Blog Content Review and Corrections

Sprint 68: Blog Content Review and Corrections

Decisions

D1: Tier classification SSoT = agents/commands/*.md

  • Context: meet-the-agents.mdx and orchestration-structure.mdx had inconsistencies in Palette's Tier and Tier 3 label ("Support" vs "Enhancement"). A reference point was needed to resolve conflicts between blog posts.
  • Choice: Adopt the [Tier N — Label] declaration on the first line of agents/commands/{name}.md as the SSoT. Blogs, documentation, and planning materials follow this value.
  • Alternatives: Adopt orchestration-structure.mdx as authoritative (blog-first) / Declare separately in MEMORY.md (memory is a snapshot, not suitable) — both rejected.
  • Code Paths: agents/commands/palette.md, agents/commands/*.md, blog/content/adr/meet-the-agents.mdx, blog/content/adr/orchestration-structure.mdx

D2: Blog series reorganization — ADR number-based → topic-based 6 posts

  • Context: Existing adr-001~005.mdx had weak narrative flow from a reader perspective due to ADR number listing. Reorganization was already underway in Sprint 67 but cross-references/Tier consistency had not been verified and were left in the working tree.
  • Choice: Finalize as 6 topic-based series and complete consistency in Sprint 68.
    1. agent-orchestration-solo-dev (intro)
    2. system-architecture-overview
    3. orchestration-structure
    4. cicd-ai-guardrails
    5. meet-the-agents
    6. sprint-journey (retrospective)
  • Alternatives: Maintain ADR number system — rejected due to poor reader navigation.
  • Code Paths: blog/content/adr/*.mdx

Patterns

P1: Document Tier/naming consistency validation is grep-based from agents/commands/

  • Where: [Tier \d declarations in agents/commands/*.md
  • When to Reuse: When writing/modifying documents describing the agent system (blog, README, ADR), first grep agents/commands/*.md to confirm the SSoT. If conflicts arise between documents, correct according to the SSoT.
  • Where: "Previous post/Next post" links in blog/content/adr/*.mdx
  • When to Reuse: When link text differs significantly from the actual post title, readers can misunderstand the next post's content. Always verify frontmatter title and link text consistency when adding/modifying series posts.

Gotchas

G1: Subagent reports are supplementary; direct grep is the final authority

  • Symptom: Explore subagent reported "Palette inconsistency in Tier 2/3, three agents → four agents correction needed" in meet-the-agents.mdx. However, Oracle's direct grep found meet-the-agents.mdx consistently described Palette as Tier 2 with "three agents". The actual error was in the overall Tier 2/3 classification.
  • Root Cause: Subagent's summary mixed observation and judgment, and Oracle nearly passed incorrect correction instructions without critical review.
  • Fix: After receiving subagent reports, Oracle directly re-verifies critical judgments with Read/Grep. Judgments like "which file is authoritative" must be confirmed directly from the SSoT.

G2: Cloudflare Tunnel routing is not visible in Ingress

  • Symptom: No blog route in kubectl get ingress during CD verification, momentarily misunderstood as "blog not exposed externally". Actually https://blog.algo-su.com was responding normally.
  • Root Cause: Blog service is delivered directly by algosu/cloudflared pod via Cloudflare Tunnel to blog Service (ClusterIP). This structure bypasses Ingress, leaving no trace in kubectl get ingress.
  • Fix: When verifying external domain routing, don't only check Ingress — also check for cloudflared pod presence + verify with actual HTTP request. Record blog domain and routing mechanism in reference_domain.md to prevent recurrence.

Metrics

  • Commits: 2 (cb86dec, 37ccf38)
  • Files changed: 16 (+2446/-571)
  • Blog post reorganization: 5 deleted + 6 new + 3 content corrections
  • Sprint ADR archive augmentation: Sprint 63~67