feat(best-practices): add scheduler expert skill#165
Open
Himanich wants to merge 10 commits into
Open
Conversation
Introduces best-practices/scheduler/ as the first expert skill subdirectory, replacing the flat references/scheduler*.md files. - SKILL.md: lean control plane — classification, path-aware review checklist (cross-cutting / Path A / Path B separated), troubleshooting fingerprints, common pitfalls, modern alternatives - path-a.md: full Path A migration steps (A1–A9), property reference table including scheduler.period/scheduler.name, configurable-cron redirect to Path B, getServiceResourceResolver null-safety note, greenfield template - path-b.md: full Path B Sling Jobs steps (B1–B5), idempotency guard pattern, canRunConcurrently() drop note, complete import lists for both Scheduler and JobConsumer classes, @Activate/@Modified dual-annotation and separate-method lifecycle variants Old reference files are deleted in a later task after all five expert skills are in place Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Himanich
requested review from
abhishekgarg18,
akankshajain18,
pkumargaddam and
rombert
as code owners
Tessl Skill Lint✅ ✅ ✅ All 2 tile(s) clean. Updated by |
… practices path-a.md: - Add resolver.commit() guidance for write-side schedulers with PersistenceException handling - Add author-only run mode guard pattern using SlingSettingsService - Add enabled flag + volatile field pattern to greenfield template - Add @Designate-backed @Modified guidance for non-scheduling config updates - Soften @deactivate checklist item to optional for stateless schedulers - Replace {best-practices} placeholder with relative path path-b.md: - Fix idempotency guard: was a no-op after unscheduleExistingJobs(); restructured as explicit alternative pattern with stale-cron limitation warning - Fix ConcurrentModificationException: wrap getScheduledJobs() in new ArrayList<>() - Add missing import java.util.ArrayList to B1 code block - Add missing import JobResult to B2 code block - Add author-only run mode guard using SlingSettingsService in B1 - Remove unused JobBuilder import from B4; replace Collection with ArrayList - Replace {best-practices} placeholder with relative path references/scheduler*.md: - Add deprecation notices pointing to scheduler/ expert skill Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
rombert
approved these changes
Jun 2, 2026
Introduces best-practices/resource-change-listener/SKILL.md as the second expert skill subdirectory, replacing the flat references/resource-change-listener.md file. Single-file SKILL.md (no split needed — one target architecture: lightweight listener + JobConsumer). Covers: - Two interface variants (ResourceChangeListener vs ExternalResourceChangeListener) - Three CS-specific constraints (no Session in onChange, service-user resolver in consumer, filter at registration) - Classification: skip / R1-R5 / R0+R1-R5 based on source type - Complete before/after example (legacy JCR EventListener → RCL + JobConsumer) - R0: convert from javax.jcr.observation.EventListener or resource-topic EventHandler - R1: keep onChange lightweight, per-change try/catch mandatory - R2: JobConsumer with resolver.commit() guidance, PersistenceException handling, author-only SlingSettingsService guard - R3: ResourceChangeListener vs ExternalResourceChangeListener decision, multi-pod fan-out warning - R4: full import diff - R5: Repoinit + ServiceUserMapperImpl.amended config - Review checklist (cross-cutting + listener + consumer + configuration) - Troubleshooting fingerprints with grep-able log strings - Common pitfalls (silent registration, topic mismatch, external-by-default, missing commit) - Modern alternatives decision matrix Cross-cutting fix (scheduler): - Remove /system/console URL references from scheduler troubleshooting; describe component runtime state instead. Felix Web Console is not accessible on AEMaaCS production — use Developer Console in Cloud Manager. Same approach applied to RCL troubleshooting. Deprecation notice added to references/resource-change-listener.md pointing to the new expert skill. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…eedback Apply reviewer feedback to bring RCL skill to expert-skill standards for AEMaaCS: - Soften absolute "all business logic must be offloaded" wording in Overview and R1; allow small in-memory operations (metrics, lightweight filtering, enqueue) in onChange() while keeping the strong rule against repository access, I/O, and expensive computation - Add AEM Eventing to Modern Alternatives as the supported path for notifying external systems (Adobe I/O Events, App Builder, webhooks); RCL clarified as in-process pattern - Add cluster-coordination warning to author-only consumer pattern: run mode check does not guarantee single execution on author clusters - Strengthen ExternalResourceChangeListener guidance with explicit pre-selection checklist (idempotency, duplicate-processing tolerance, intentional per-node execution) - Soften "not recommended" wording for javax.jcr.observation.EventListener; preserve escalation path (Adobe support case for documented platform needs) - Add PROPERTY_NAMES_HINT semantics note: optimization hint, not a contract; application code must validate actual ResourceChange data - Add Common Pitfalls: event storms from broad PATHS, recursive self-triggering when consumer writes to observed subtree - Add Troubleshooting rows: repeated self-triggering, excessive job volume - Add Expert Guidance section: idempotency principles for distributed AEMaaCS (event batching, retries, cluster coordination, derive state from repository) - Add note that PATHS are chosen by the implementer from the business scope; do not ask customer for raw PATHS syntax Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
mandeep294
pushed a commit
to mandeep294/skills
that referenced
this pull request
Jun 3, 2026
Design doc for merging health-assessment into code-assessment as a single code-quality skill with per-pattern expert skills (PR adobe#165 shape). MCP becomes an optional accelerator with auto-fallback to local discovery; one shared engine owns all repo-touching logic. Captures locked decisions (Q1 fallback, Q2 per-pattern resolution contract, Q3 single engine, Q5 report-all/apply-one) and carries over shared-principles unchanged. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
mandeep294
pushed a commit
to mandeep294/skills
that referenced
this pull request
Jun 3, 2026
8-task plan to merge health-assessment into code-assessment as a single MCP-optional skill with per-pattern expert skills (PR adobe#165 shape). Each task ends in skills-ref validate + dangling-reference grep + commit. No manifest edits (skills auto-discovered). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
mandeep294
pushed a commit
to mandeep294/skills
that referenced
this pull request
Jun 3, 2026
Task 1 of the umbrella-skill plan. Moves _template.md into code-assessment/ references/ and rewrites it to the per-pattern expert-skill shape (PR adobe#165). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Introduces best-practices/replication/SKILL.md as the third expert skill subdirectory, replacing the flat references/replication.md file. Single-file SKILL.md covering the Sling Distribution API (Distributor + SimpleDistributionRequest) end-to-end: - Migration from CQ Replicator (com.day.cq.replication.Replicator) and Sling Replication Agent (org.apache.sling.replication.agent.api) to the Sling Distribution API - Predefined agents on AEMaaCS: publish (default) + preview (opt-in) - ReplicationActionType → DistributionRequestType mapping table - P1-P4 transformation steps with imports and service-user setup - Repoinit + ServiceUserMapperImpl.amended with principal-name mapping; crx:replicate privilege requirement - Author-only execution guard + author cluster coordination warning - Greenfield template: agent iteration, defensive preview handling, isAgentNotConfigured helper, JobResult-style outcome classification - Composition example: ResourceChangeListener → JobConsumer → Distributor with DistributionResponse → JobResult mapping (CANCEL for unrecoverable permission/agent errors; FAILED for transient queue/transport) - Tree Activation workflow as Adobe's recommended path for bulk publishing (not custom loops) - Distribution UI guidance: Tools > Deployment > Distribution, queue states (persisted → fully published), Queue items and Logs panels, blocked/paused agent status - Async semantics emphasis: isSuccessful() means queued, not delivered - Troubleshooting fingerprints with grep-able log strings (state=DROPPED, not authorized, no matching agent, ClassNotFoundException) - Common pitfalls (legacy agent names, deep=true on broad paths, USER/PASSWORD auth maps, treating async as sync, building custom bulk loops, swallowing DistributionException) - Expert Guidance with explicit Practical Engineering Notes subsection separating Adobe-documented behavior from engineering observations (de-duping, OSGi event topics) - Modern Alternatives matrix including AEM Eventing for external notification Deprecation notice added to references/replication.md pointing to the new expert skill. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…e-listener feat(best-practices): add resource-change-listener expert skill
feat(best-practices): add replication expert skill
Introduces best-practices/event-migration/SKILL.md as the fourth expert skill subdirectory, replacing the flat references/event-migration*.md files (router + path-a + path-b consolidated into one expert skill). Single-file SKILL.md (~800 lines) covering the lightweight EventHandler + JobConsumer pattern for non-resource OSGi Event Admin topics: - Routing-first rules: resource events → resource-change-listener skill, external notification → AEM Eventing, only non-resource OSGi topics (replication, workflow, custom) belong in this skill - Topic discovery guidance: canonical topic constants (ReplicationEvent, WorkflowEvent, NotificationConstants) and warning that typo'd topic strings silently subscribe to nothing - Three CS-specific constraints: no JCR ops in handleEvent, service-user resolver in consumer, TopologyEventListener for leader-only execution - Migration steps E0-E5 covering legacy JCR EventListener (rare) and OSGi EventHandler with inline logic - Complete before/after example: ReplicationEvent handler with isLeader gate + separate JobConsumer with resolver.commit() and PersistenceException handling - Workflow event variant: shows only the deltas from the replication example (topic, event-type discriminator, WorkflowEvent payload keys) - Multi-topic / wildcard subscription pattern with event.getTopic() dispatch - TopologyEventListener pattern: when to add leader election, when to skip, volatile field requirement, TOPOLOGY_INIT and TOPOLOGY_CHANGED handling - Greenfield rules and JobConsumer template with JobResult outcome classification (CANCEL for missing required data, FAILED for transient, OK for success) - Composition: EventHandler → JobConsumer → Distributor cross-reference to replication skill - Repoinit + ServiceUserMapperImpl.amended with principal-name mapping - Review checklist: cross-cutting + EventHandler + leader election + JobConsumer + configuration - Troubleshooting fingerprints with grep-able log strings - Common pitfalls: business logic in handleEvent, deprecated resource topics, run-mode as singleton guard, missing volatile on isLeader, topic constant mismatch, OSGi for external notification, TopologyEventListener without consuming topology - Modern Alternatives with AEM Eventing for external systems - Expert Guidance with idempotency principles for distributed AEMaaCS - Practical Engineering Notes separating Adobe-documented behavior from engineering observations (TOPOLOGY_CHANGING reset, event ordering, LDAP filter syntax) Deprecation notices added to references/event-migration.md, event-migration-path-a.md, and event-migration-path-b.md pointing to the new expert skill. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Reverts the size-driven trims to the workflow variant, Greenfield template, and Composition section. The trims removed redundancy but forced the LLM to mentally reconnect patterns shown elsewhere, which is the actual hallucination risk vector — not raw file size. Restored: - Workflow variant: full standalone EventHandler class (was: deltas-only). LLM can now copy-paste a workflow handler without reassembling from the replication After example. - Greenfield template: full File 1 (EventHandler + TopologyEventListener) + File 2 (JobConsumer with outcome classification) (was: JobConsumer + rules only). LLM has a complete handler+consumer pair to pattern-match against. - Composition section: full process() method with try-with-resources and exception handling (was: just the integration call). LLM gets the full pattern for EventHandler -> JobConsumer -> Distributor composition. Trade-off accepted: file is back to ~930 lines (vs RCL 662 / replication 728). Event-migration legitimately covers more (E0-E5 + workflow variant + composition + leader election). Self-contained examples reduce pattern-reconstruction errors when the LLM generates code for any of the three use cases, which matters more than file length at this scale (Claude 4.x handles 900+ line skills comfortably). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Feat/best practices event migration
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Introduces best-practices/scheduler/ as the first expert skill subdirectory, replacing the flat references/scheduler*.md files.
after all five expert skills are in place
Description
Related Issue
Motivation and Context
How Has This Been Tested?
Screenshots (if appropriate):
Types of changes
Checklist: