feat(aem-cloud-service): add guava-cache analyzer detector + expert skill by bharat941 · Pull Request #331 · adobe/skills

bharat941 · 2026-09-04T05:51:59Z

Adds the guava-cache code-assessment pattern — Guava cache (com.google.common.cache.*) → Caffeine (com.github.benmanes.caffeine.cache.*) on AEM as a Cloud Service. Follows the references/adding-a-pattern.md procedure; the slug guava-cache is wired into all four required places so the [wiring] test stays green.

Changes

Detector — scripts/analyzer/detectors/GuavaCache.java, registered in Registry.all(). Import-anchored on com.google.common.cache.* (Java-only). Exact package prefix means the io.micrometer…GuavaCacheMetrics BPA false positive is not flagged.
Expert skill — guava-cache/SKILL.md (control plane) + guava-cache/recipe.md (C1 pom swap, C2 imports, C3 builder/API mapping incl. getUnchecked→get, Callable→Function, RemovalNotification→3-arg; test generation).
Catalog + routing — references/patterns.md row (low | ready | analyzer | guided) and Manual Pattern Hints row in code-assessment/SKILL.md.
Fixtures + test — test/code-assessment/fixtures/guava-cache/ (legacy guava / clean caffeine / micrometer guard) + a block in run-tests.sh.

Test Plan

run-tests.sh — 109 PASS / 0 FAIL (incl. [wiring] and the new [guava-cache] block). Detector flags com.google.common.cache.* imports; does not flag Caffeine or GuavaCacheMetrics. End-to-end on a real legacy AEM project: analyzer detects → recipe transform applied → analyzer re-scan clean → mvn compile BUILD SUCCESS.

Supersedes #326 — re-raised from a branch on adobe/skills now that write access is available (identical commits).

…kill Adds the guava-cache code-assessment pattern (Guava cache -> Caffeine on AEM CS): - GuavaCache detector: import-anchored on com.google.common.cache.* (Java-only), registered in Registry.all(); exact prefix avoids the micrometer GuavaCacheMetrics false positive. - guava-cache/ expert skill: SKILL.md (control plane) + recipe.md (C1 pom swap, C2 imports, C3 builder/API mapping, test generation). - Catalog row in patterns.md (low | ready | analyzer | guided) + Manual Pattern Hints routing row in code-assessment SKILL.md. - Fixtures (legacy guava / clean caffeine / micrometer guard) + run-tests.sh block. Analyzer suite: 109 PASS / 0 FAIL (incl. [wiring] + guava block).

github-actions · 2026-09-04T05:52:18Z

Tessl Plugin Lint

⚠️

edge-delivery-services — 2 warning(s)

⚠ Skill 'code-review': SKILL.md is approximately 10352 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.
⚠ Skill 'figma-to-content': SKILL.md is approximately 14744 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.

✔ Plugin adobe/aem-edge-delivery-services@0.1.0 is valid

⚠️

app-management — 1 warning(s)

⚠ Skill 'commerce-app-admin-ui': SKILL.md is approximately 5265 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.

✔ Plugin adobe/commerce-app-management@1.3.2 is valid

✅ app-migration — clean

✅ app-review — clean

✅ run-workflow — clean

⚠️

stardust — 9 warning(s)

⚠ Skill 'extract': SKILL.md is approximately 9459 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.
⚠ Skill 'direct': SKILL.md is approximately 11694 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.
⚠ Skill 'prototype': SKILL.md is approximately 17875 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.
⚠ Skill 'migrate': SKILL.md is approximately 6375 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.
⚠ Skill 'uplift': SKILL.md is approximately 6515 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.
⚠ Skill 'audit': SKILL.md is approximately 5269 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.
⚠ Skill 'reskin': SKILL.md is approximately 5906 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.
⚠ Skill 'deploy': SKILL.md is approximately 38193 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.
⚠ Skill 'rollout': SKILL.md is approximately 6779 tokens (recommended maximum: 5000). Consider moving detailed content to separate reference files.

✔ Plugin adobe/stardust@0.18.1 is valid

✅ All 6 plugin(s) lint passed with 12 warning(s) total.

Updated by tessl-lint for commit cfaaf71.

Himanich · 2026-09-04T08:02:53Z

 | [`asset-manager`](../asset-manager/SKILL.md) | migrate DAM `AssetManager` create/upload via Direct Binary Access (`@adobe/aem-upload`) and delete via in-JVM `resolver.delete()` + `commit()` or HTTP Assets API; removes `createAssetForBinary` / `getAssetForBinary` / `removeAssetForBinary` (not available on CS) | high | ready | analyzer | guided |
 | [`outbound-call-timeouts`](../outbound-call-timeouts/SKILL.md) | add connect/read/socket timeouts to outbound HTTP client construction (Apache HttpClient, OkHttp, JDK HttpClient) | high | ready | analyzer | mechanical |
 | [`unbounded-query`](../unbounded-query/SKILL.md) | bound or escalate an explicitly-unbounded query (`p.limit=-1` predicate / `setLimit(-1)`) — safe-cap where provable, else flag for pagination | high | ready | analyzer | guided |
+| [`guava-cache`](../guava-cache/SKILL.md) | swap Guava cache (`com.google.common.cache.*`) for Caffeine (`com.github.benmanes.caffeine.cache.*`) — pom dependency + imports + API call sites | low | ready | analyzer | guided |


fix: guided contradicts this pattern's self-evident contract (SKILL.md:37) and "mechanical" Overview (SKILL.md:13). Every other pattern keeps fix and contract aligned — guided↔guided, mechanical↔self-evident. Runbook.md:199 renders guided as apply (guided), mis-dispositioning these deterministic edits. Make it mechanical (like outbound-call-timeouts), or switch the contract to guided.

Also, severity: low (patterns.md:42) is the only low in the catalog — every other pattern is medium/high.

Also, follow the lui/cdw/htlLint precedent: drop it from Registry.java, patterns.md, and Hints; add a routing row in migration/SKILL.md with the recipe under migration/references/, discovered via BPA.

Himanich · 2026-09-04T08:03:17Z

+
+## Resolution contract
+
+**self-evident** — the Guava → Caffeine API mapping is fixed (see [recipe.md](recipe.md)); no user input is required to plan the edit. The only judgment is the Caffeine version: pin it to the AEM CS SDK BOM (default `3.1.8`).


self-evident conflicts with fix: guided in patterns.md:42 — reconcile the two.

Himanich · 2026-09-04T08:31:17Z

+
+## Discovery
+
+Detection is performed by the analyzer ([`../scripts/analyze.sh`](../scripts/README.md)), run by the runbook:


This finding already comes from BPA, so discovery should match the other migration patterns: BPA report (MCP or local CSV) first → local scan only as fallback. As wired (analyzer detector + ready|analyzer row), the sole path is a local scan that ignores the existing BPA finding. Source it via getBpaFindings first, local scan as fallback.

…ation Guava cache usage doesn't occur in native AEMaaCS code, only in code carried over from legacy AEM — so per review feedback this belongs to the migration skill's domain, not code-assessment's. - Remove the guava-cache expert skill, analyzer detector, and fixtures from code-assessment; revert its Registry/SKILL.md/patterns.md/run-tests.sh hooks. - Add migration/references/guava-cache.md (BPA-driven reference, no dedicated pattern guide — same shape as htlLint). - Wire guavaCache as a BPA-only pattern (subtype com.google.common.cache) into bpa-local-parser.js, unified-collection-reader.js, and runbook-generator.js's PATTERN_META — one finding per file, not per import, since BPA already reports at file granularity. No analyzer/content-scan fallback is added; with no BPA source the pattern surfaces via the existing LLM-scan tier. - Drop internal "avoids the known BPA false positive" language; BPA is now the stated source of truth for this pattern. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

…ummary fetch-cam-bpa-findings-by-pattern's pattern enum already includes guavaCache server-side; the migration skill's own summary of that tool's patterns was missing it. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Verified against a real customer BPA report (64k rows): the actual subtype is `custom.guava.cache`, not `com.google.common.cache` as assumed from the reference doc alone. More importantly, `identifier` on this subtype is a Guava-internal class (e.g. com.google.common.cache.AbstractCache) reachable on a bundle's classpath, not the customer's own class — BPA bytecode-scans Guava's own cache implementation wherever it's embedded, so one real bundle produced 1591 raw rows for two actual bundles. - Group by bundle name (extracted from the message text), not by `identifier` — dedupes to one finding per bundle instead of ~800/bundle. - Fix the subtype string everywhere it's checked/mapped. - Update the reference doc and migration/SKILL.md to describe bundle-level discovery instead of file-level, and to explain why `identifier` can't be used directly. Confirmed end-to-end against the real 64k-row report: getBpaFindings('guavaCache', ...) now correctly returns exactly the 2 real bundles instead of ~1591 bogus entries. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

bharat941 requested review from Himanich, abhishekgarg18, akankshajain18, pkumargaddam and rombert as code owners September 4, 2026 05:52

bharat941 requested a deployment to eval September 4, 2026 05:52 — with GitHub Actions Waiting

bharat941 mentioned this pull request Sep 4, 2026

feat(aem-cloud-service): add guava-cache analyzer detector + expert skill #326

Closed

Himanich reviewed Sep 4, 2026

View reviewed changes

bharat941 requested a deployment to eval September 4, 2026 10:21 — with GitHub Actions Waiting

bharat941 requested a deployment to eval September 4, 2026 10:23 — with GitHub Actions Waiting

bharat941 requested a deployment to eval September 4, 2026 10:34 — with GitHub Actions Waiting

Navigation Menu

Uh oh!

feat(aem-cloud-service): add guava-cache analyzer detector + expert skill - #331

feat(aem-cloud-service): add guava-cache analyzer detector + expert skill#331
bharat941 wants to merge 4 commits into
mainfrom
feat/guava-cache-code-assessment

bharat941 commented Sep 4, 2026

Uh oh!

github-actions Bot commented Sep 4, 2026 •

edited

Loading

Uh oh!

Himanich Sep 4, 2026

Uh oh!

Himanich Sep 4, 2026

Uh oh!

Himanich Sep 4, 2026

Uh oh!

Himanich Sep 4, 2026

Uh oh!

Himanich Sep 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants


		## Resolution contract

		self-evident — the Guava → Caffeine API mapping is fixed (see [recipe.md](recipe.md)); no user input is required to plan the edit. The only judgment is the Caffeine version: pin it to the AEM CS SDK BOM (default `3.1.8`).


		## Discovery

		Detection is performed by the analyzer ([`../scripts/analyze.sh`](../scripts/README.md)), run by the runbook:

Uh oh!

Conversation

bharat941 commented Sep 4, 2026

Changes

Test Plan

Uh oh!

github-actions Bot commented Sep 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Tessl Plugin Lint

Uh oh!

Himanich Sep 4, 2026

Choose a reason for hiding this comment

Uh oh!

Himanich Sep 4, 2026

Choose a reason for hiding this comment

Uh oh!

Himanich Sep 4, 2026

Choose a reason for hiding this comment

Uh oh!

Himanich Sep 4, 2026

Choose a reason for hiding this comment

Uh oh!

Himanich Sep 4, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

github-actions Bot commented Sep 4, 2026 •

edited

Loading