corp-web-japan-publication-gap-audit

name	corp-web-japan-publication-gap-audit
description	Audit missing publication IDs in corp-web-japan and decide whether each gap should become a hidden redirect record, a full local migration, a redirect-policy fix, or no action.
version	1.0.0
author	Hermes Agent
license	MIT
metadata	{"hermes":{"tags":["corp-web-japan","publications","id-gaps","hidden-posting","redirect","corp-web-contents","vercel-logs"]}}

corp-web-japan publication ID gap audit

Use this when reviewing corp-web-japan publication/content routing and you notice missing numeric IDs in local content directories such as:

• src/content/whitepapers/*.mdx
• src/content/events/*.mdx
• src/content/news/*.mdx
• src/content/blog/*.mdx
• src/content/demo/**
• src/content/use-cases/*.mdx

Why this skill exists

A missing ID does not automatically mean:

• a hidden posting should be added
• a redirect-only shadow record should be created
• local migration is missing

The correct answer depends on:

1. whether the source ever existed in ../corp-web-contents
2. whether a Japanese (ja) source exists there now
3. whether production traffic still hits the missing legacy/current paths
4. whether the gap is already intentionally locked in by repo tests

In practice, the user expects this gap/source/traffic audit before a conclusion like “this PR is stale” or “just add hidden + redirect”.

Core rule

Do not jump directly from “ID gap” to “create hidden posting”.

Always classify the gap first:

• full-local-migration candidate
• hidden redirect record candidate
• legacy redirect-policy candidate
• separate anomaly / needs more source tracing
• intentional omission / no action

Recommended workflow

1. Inventory local ID gaps first

From the repo root, enumerate actual IDs and gaps per family.

Example:

python3 - <<'PY'
import os,re,glob
roots=[
 ('blog','src/content/blog'),
 ('whitepapers','src/content/whitepapers'),
 ('news','src/content/news'),
 ('events','src/content/events'),
 ('use-cases','src/content/use-cases'),
 ('demo-aip','src/content/demo/aip'),
 ('demo-acp','src/content/demo/acp'),
]
for name,root in roots:
    ids=sorted(int(os.path.splitext(os.path.basename(p))[0]) for p in glob.glob(root+'/*.mdx') if re.match(r'^\d+\.mdx$', os.path.basename(p)))
    gaps=[]
    if ids:
        gaps=sorted(set(range(min(ids), max(ids)+1))-set(ids))
    print(f'[{name}] count={len(ids)} gaps={gaps}')
PY

Also list actual filenames to avoid bad assumptions.

2. Check whether tests already encode the omission intentionally

Before deciding a gap is accidental, inspect corpus tests such as:

• tests/whitepaper-imported-ja-corpus.test.mjs
• tests/events-imported-ja-corpus.test.mjs
• family-specific imported-corpus tests

If the missing IDs are explicitly excluded from expectedIds, treat the omission as intentional current behavior, even if it may still deserve reconsideration.

Important interpretation:

• test omission means “currently expected state”
• it does not prove the omission is correct product policy

3. Investigate ../corp-web-contents source existence

Stay grounded in the sibling repo the user named.

For each missing ID, inspect current/history paths.

Useful commands:

cd ../corp-web-contents

# broad path discovery by id
for id in 13 14; do
  git log --all --name-only --pretty=format: | sed '/^$/d' | sort -u | grep "/$id/" || true
done

Prefer narrowing by family patterns such as:

• /white-paper/<id>/
• /webinars/<id>/
• /blog/<id>/
• /use-cases/<id>/

4. Distinguish “source exists” from “JA migration candidate exists”

This is the critical fork.

A. Current canonical JA source exists in corp-web-contents

Example shape:

• pages/features/documentation/white-paper/<id>/<slug>/ja/content.mdx

Interpretation:

• the gap is likely a full local migration candidate
• do not default to a hidden redirect record first
• ask why the local corpus skipped migration of a still-existing JA source

B. Only EN/KO exists, but no JA canonical source exists

Example shape:

• pages/features/demo/webinars/<id>/<slug>/en/content.mdx
• pages/features/demo/webinars/<id>/<slug>/ko/content.mdx
• no ja/content.mdx

Interpretation:

• this is not a strong local Japanese migration candidate
• prefer classifying it as a legacy redirect-policy candidate rather than creating a local Japanese posting shell just to fill the number

5. Check production traffic before deciding priority

Use Vercel production logs for live evidence.

Fast existence check:

vercel logs --project corp-web-japan --environment production --since 24h --json --no-branch --limit 20

Then query specific suspected paths:

vercel logs --project corp-web-japan --environment production --since 30d --search '/features/documentation/white-paper/13' --json --no-branch --limit 20
vercel logs --project corp-web-japan --environment production --since 30d --search '/features/demo/webinars/7' --json --no-branch --limit 20
vercel logs --project corp-web-japan --environment production --since 30d --search '/events/6' --json --no-branch --limit 20

Important patterns:

• runtime-missing-redirect + 307 means legacy traffic still hits the path
• a direct 404 on a would-be current canonical path means the gap may already be user-visible

6. Classification rules

Class 1: full local migration candidate

Use this when:

• missing local ID exists in corp-web-contents current canonical source
• current canonical ja/content.mdx exists there
• legacy traffic may also exist

Typical example:

• whitepaper IDs missing locally even though ../corp-web-contents/pages/features/documentation/white-paper/<id>/.../ja/content.mdx exists

Recommendation:

• prioritize re-evaluating why the local migration skipped this ID
• hidden redirect record is only a fallback, not the first assumption

Class 2: hidden redirect record candidate

Use this when:

• the gap corresponds to an old/local detail URL that should keep resolving
• local list visibility should stay off
• a stable target exists
• creating a full local migration is unnecessary or undesirable

Important note:

• choose this only after rejecting full migration as the more correct path

Class 3: legacy redirect-policy candidate

Use this when:

• historical source exists
• production legacy path traffic exists
• but there is no current Japanese canonical source to migrate

Typical example:

• old webinar IDs with EN/KO only, no JA source, but bots/users still hit /features/demo/webinars/<id>/...

Recommendation:

• review allowlist / redirect behavior first
• do not force-fit a local Japanese posting purely for numeric continuity

Class 4: separate anomaly

Use this when:

• production logs show a hit to a current-style path such as /events/6/<slug>
• but the slug does not match the obvious legacy source lineage
• and corp-web-contents history does not immediately explain it

Recommendation:

• trace the slug lineage separately before proposing hidden/redirect/migration

Class 5: intentional omission / no action for now

Use this when:

• repo tests intentionally exclude the ID
• no corp-web-contents source exists
• no production traffic exists

7. Report format

For each family, summarize like this:

Family: whitepapers
Gap IDs: 13, 14

ID 13
- corp-web-contents canonical JA source: yes
- production traffic evidence: yes (`/features/documentation/white-paper/13/...` 307)
- classification: full local migration candidate
- note: do not jump straight to hidden redirect

Practical findings that prompted this skill

Whitepaper gap pattern

• local src/content/whitepapers can omit IDs that still exist in ../corp-web-contents/pages/features/documentation/white-paper/<id>/.../ja/content.mdx
• if production logs also show requests to those legacy feature paths, the omission deserves migration review, not an automatic hidden shadow record

Event/webinar gap pattern

• local src/content/events can intentionally omit webinar IDs that exist in corp-web-contents only as EN/KO source without ja/content.mdx
• if production logs show runtime-missing-redirect traffic to /features/demo/webinars/<id>/..., that usually points to redirect-policy work rather than local Japanese MDX migration

Test-encoded omission pattern

• imported-corpus tests may explicitly freeze current omitted IDs in expectedIds
• this means the gap is deliberate in current repo state, but not necessarily the correct end state

Pitfalls

• Assuming every gap should be filled with a local file
• Assuming every missing ID should be hidden + redirect
• Ignoring corpus tests that intentionally freeze omitted IDs
• Treating historical source existence as enough for local migration without checking JA source availability
• Treating legacy-path traffic as proof that a local canonical page is required; often it only proves redirect-policy relevance
• Reviewing a publication/routing PR only for stale/mergeable status and missing the more important hidden/redirect/gap question

Done criteria

A gap audit is complete when you can answer, for each missing ID:

• did a source exist in corp-web-contents?
• does a canonical JA source exist there now?
• is there production traffic evidence?
• is the omission explicitly encoded in tests?
• should this become full migration, hidden redirect, redirect-only policy work, or no action?