Skip to content

docs-site: remove ADR references from the intro deck index - #195

Closed
Atom-oh wants to merge 2 commits into
devfrom
docs-site/deck-index-no-adr
Closed

Atom-oh wants to merge 2 commits into
devfrom
docs-site/deck-index-no-adr

Conversation

@Atom-oh

@Atom-oh Atom-oh commented Oct 22, 2026 •

Copy link
Copy Markdown
Contributor

Superseded — do not merge. The combined index.html edits now ride with their source blocks in #196, #197 and #205 (each carries its Markdown source, per-part rendering and matching index.html lines). Once those three merge, index.html on dev equals this branch's version and this PR can be closed.

🤖 Generated with Claude Code

Split from #192 to stay under the AI review input limit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@github-actions

github-actions Bot commented Oct 22, 2026 •

Copy link
Copy Markdown

🤖 AI Code Review (two independent reviewers)

_Reviewer responses: codex/ALL claude/ALL _

Status: BLOCKED — CRITICAL/MAJOR issues must be resolved before merge

All panel claims are now verified directly against the base checkout. Writing the chair synthesis.

Chair Review — PR #195: docs-site: remove ADR references from the intro deck index

1. Summary

The PR hand-edits the combined web deck docs-site/static/presentation/awsops-intro/index.html to strip all 42 ADR citations (ADR-001/002/003/004/005/007/009/011/013/015/030) from slide bodies and the inline presenter-notes object. The edits are textual only — no code, IaC, or policy surface is touched, and the substance of the read-only/FROZEN posture is preserved in the rewritten prose. However, the file is a generated artifact whose declared sources (_presentation.md + 0N-*.md blocks) and sibling renderings (0N-*.html) are untouched and still carry every removed string verbatim, so the change is both self-reverting on regeneration and internally inconsistent within the published deck directory.

2. Issues per lens

L2 — Code correctness

  • No blocking issues (both panels agree). I spot-verified the diff's mechanics: only string/text-node content changes; object keys of the notes literal are unchanged; the one inner double quote (\"read-only\"는 in note 10) remains escaped; no tags or data-remarp-id attributes altered. The notes map still parses and slide indexing is unaffected.
  • MINOR (claude only): awkward phrasing introduced in notes 36 ("…동결 원칙에 따라 동결됐습니다" — "frozen per the freeze principle") and 50. Cosmetic.

L3 — Security / AWS mutation safety

  • No blocking issues (both panels agree). No IAM/Terraform/agent path is touched; the ADR-005 freeze, the single ecs:UpdateService force-new-deployment exception with its full constraint list (secret-rotation trigger, one ARN, fail-closed, default-off), the remediation_enabled FROZEN marker, and the governed-external-write description all survive the rewrite. No secrets/ARNs/account IDs introduced.
  • MINOR (claude only, verified): the Gated Subsystems row for secret-rotation.tf loses its only governance marker ((ADR-015) → nothing), leaving a mutation-capable subsystem listed with no indication it is the sole sanctioned freeze exception. A non-numeric marker ("승인된 동결 예외 1건 · default-off") would preserve the signal.

L4 — Observability / data-integration correctness

  • No blocking issues (both panels agree). No Steampipe SQL, CIS/Powerpipe, collector, or diagnosis logic in scope. The two re-worded factual claims (hard-false steampipeAvailable() with fail-open to ops; diagnosis_digest scope caveat) remain accurate against the repo's documented behavior.
  • MINOR (claude only): "ADR-001/010에 따라" → "설계 원칙에 따라" removes the only traceability pointer explaining why the dead-collector path is intentional.

L5 — Docs / ADR consistency

The panels disagree outright here: codex reported no findings; claude reported three MAJORs. I verified claude's claims directly and codex is wrong on two of them — this is not a shared-bias convergence case, since the evidence is in the base files:

  • MAJOR (confirmed) — generated artifact edited, sources untouched. awsops-intro/README.md explicitly declares index.html + 0N-*.html as generated from the _presentation.md manifest + 0N-*.md blocks. The source blocks still contain every removed string verbatim (e.g. 02-architecture.md:114 remediation_enabled — **ADR-005 FROZEN**, :357 # Hybrid Routing & Streaming Chat — ADR-003, :772 the dated ADR-015 owner-override block, :805). Failure trigger: the next regeneration from the declared source of truth silently reverts all 25 slide edits. The fix belongs in the three .md blocks, with the HTML regenerated in the same commit.
  • MAJOR (confirmed) — intra-deck divergence ships from one directory. The per-part decks published alongside index.html retain their ADR references (01-why-awsops.html: 7, 02-architecture.html: 25 — e.g. :184 slide header still reads "… — ADR-003", 03-demo-scenarios.html: 10; totals exactly match the 42 removed from the combined deck). The same slide now renders with ADR citations in one file and without in another. The removal must cover sources + all four HTML renderings, or not land.
  • MINOR (downgraded from claude's third MAJOR): root CLAUDE.md sanctions citing ADRs by number "for traceability only," but it does not mandate citations in customer-facing docs-site material — scrubbing internal decision-record numbers from a public product deck is a defensible editorial choice, so this is not independently gate-blocking. The actionable residue (inconsistent application, no documented convention) is already captured by the two MAJORs above.
  • MINOR (confirmed): the separate PPTX deck (build-awsops-intro-pptx.js, 7 ADR refs) was not refreshed; the deck README asks for the PPTX to be updated when web-slide content changes materially, and no CI gate checks HTML↔Markdown parity, so the divergence is silent.
  • MINOR: the exception callout loses its date (2026-07-01 owner-override) and the "나머지 ADR-005 전부" antecedent — policy substance preserved, provenance weakened. Keeping the date (while dropping the number/name) would retain auditability.

3. Suggestions

  1. Re-do the change at the source layer: edit 01-why-awsops.md / 02-architecture.md / 03-demo-scenarios.md (and _presentation.md if headers are listed there), regenerate index.html and the three 0N-*.html part decks in the same commit.
  2. Rebuild awsops-intro.pptx via node scripts/pptx/build-awsops-intro-pptx.js in the same change (its provenance gate requires the script-built binary anyway), stripping the 7 ADR refs from the build script.
  3. Where an ADR number carried a governance signal, substitute a non-numeric marker rather than deleting the clause: keep the 2026-07-01 date on the sole freeze exception, and tag the secret-rotation.tf row as the approved exception / default-off.
  4. Record the new convention ("no internal ADR identifiers in publicly deployed deck material") in the deck README so the next regeneration or deck edit doesn't reintroduce them.

4. Verdict

Two confirmed MAJOR L5 findings (generated-artifact edit that regeneration will revert; published intra-deck divergence). L2/L3/L4 are clean.

IMAGE_COVERAGE: NOT_REQUIRED


Triggered by commit 36de90e34a9ce0db4fd6a853f3c605f5e23138a5 · workflow: .github/workflows/pr-review.yml

Keep the facts the ADR removal dropped: the diagnosis digest's approved
scope is the scheduled summary while the implementation also batches
manual runs, and lifting the AWS-mutation freeze needs a new decision
document, a multi-AI panel review and a dated owner approval.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Atom-oh added a commit that referenced this pull request Oct 24, 2026
… PRs

The architecture section (#196) and the regenerated combined index.html
(#195, merged last) carry their own changes, so each review stays under
the input limit with its sources and renderings together.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Atom-oh added a commit that referenced this pull request Oct 24, 2026
index.html is regenerated from the part sources in #195 after this
section and the other parts merge, so this review covers the
architecture source and its rendering together.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@Atom-oh

Atom-oh commented Oct 24, 2026

Copy link
Copy Markdown
Contributor Author

Superseded by #196, #197 and #205: each carries its own source .md, per-part HTML and matching index.html lines, so the combined deck stays consistent at every merge. Once all three merge, dev's index.html equals this branch's version.

@Atom-oh Atom-oh closed this Oct 24, 2026
Atom-oh added a commit that referenced this pull request Oct 24, 2026
)

* docs-site: remove ADR references from the intro deck index

Split from #192 to stay under the AI review input limit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs-site: remove ADR references from the deck architecture section

Split from #192 to stay under the AI review input limit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs-site: remove ADR references from the deck sources and builder

Regenerates awsops-intro.pptx with scripts/pptx/build-awsops-intro-pptx.js
(verify-deck.sh passes); HTML is edited in step with its Markdown source.

Split from #192 to stay under the AI review input limit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix: address review feedback (iteration 1/5)

Keep the facts the ADR removal dropped: the diagnosis digest's approved
scope is the scheduled summary while the implementation also batches
manual runs, and lifting the AWS-mutation freeze needs a new decision
document, a multi-AI panel review and a dated owner approval.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix: address review feedback (iteration 1/5)

Keep the facts the ADR removal dropped: the diagnosis digest's approved
scope is the scheduled summary while the implementation also batches
manual runs, and lifting the AWS-mutation freeze needs a new decision
document, a multi-AI panel review and a dated owner approval.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix: address review feedback (iteration 1/5)

Keep the facts the ADR removal dropped: the diagnosis digest's approved
scope is the scheduled summary while the implementation also batches
manual runs, and lifting the AWS-mutation freeze needs a new decision
document, a multi-AI panel review and a dated owner approval.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs-site: leave the architecture part and combined deck to their own PRs

The architecture section (#196) and the regenerated combined index.html
(#195, merged last) carry their own changes, so each review stays under
the input limit with its sources and renderings together.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore: request review after retargeting to dev

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs-site: carry the why/demo slide edits into the combined deck

index.html renders the same blocks, so the why-awsops and demo-scenario
slide and speaker-note lines change with their sources in this PR.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Atom-oh added a commit that referenced this pull request Oct 24, 2026
…slides 1-9) (#196)

* docs-site: remove ADR references from the intro deck index

Split from #192 to stay under the AI review input limit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs-site: remove ADR references from the deck architecture section

Split from #192 to stay under the AI review input limit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix: address review feedback (iteration 1/5)

Keep the facts the ADR removal dropped: the diagnosis digest's approved
scope is the scheduled summary while the implementation also batches
manual runs, and lifting the AWS-mutation freeze needs a new decision
document, a multi-AI panel review and a dated owner approval.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix: address review feedback (iteration 1/5)

Keep the facts the ADR removal dropped: the diagnosis digest's approved
scope is the scheduled summary while the implementation also batches
manual runs, and lifting the AWS-mutation freeze needs a new decision
document, a multi-AI panel review and a dated owner approval.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs-site: leave the combined deck to its own PR

index.html is regenerated from the part sources in #195 after this
section and the other parts merge, so this review covers the
architecture source and its rendering together.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* chore: request review after retargeting to dev

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs-site: limit this part to architecture slides 1-9 and their combined-deck lines

Slides 10-19 move to a sibling PR so each review stays under the input
limit; the combined index.html lines for slides 1-9 change here with
their source block and per-part rendering.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

This branch had an error being deployed

1 failed deployment
ci-review-auto — 36de90e3 Deployed Oct 22, 2026 by Atom-oh via AI Code Review #773
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant