Skip to content

CHANGELOG.md

Latest commit

 

History

History
527 lines (413 loc) · 34.2 KB

CHANGELOG.md

File metadata and controls

527 lines (413 loc) · 34.2 KB
Raw

Changelog

All notable changes to SAP Skills will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

Security

  • Hardened all 8 GitHub Actions workflows against supply-chain attacks: pinned all 20 action references to verified commit SHAs with persist-credentials: false on all 9 checkout steps, declared least-privilege permissions: plus per-job timeout-minutes and concurrency groups in every workflow, removed all ${{ }} template expressions from run: shell bodies (values now arrive through step env: mappings), and disabled npm package-manager caching in the release-signing workflow (package-manager-cache: false) to close the cache-poisoning exposure.
  • Added a workflow lint gate to quality-checks.yml (npm run lint:workflows: actionlint v1.7.12 + zizmor v1.29.0, installed in CI from pinned releases verified against embedded SHA-256 digests) that fails on any finding; the gate is green with zero suppression annotations. Background, incident research, and watch items: docs/project/github-actions-security-hardening-2026-08-29.md.

[2.4.1] - 2026-08-07

Changed

  • Bumped the coordinated repository, marketplace, plugin manifest, and skill metadata versions to 2.4.1.
  • Updated 4 drifted MCP packages to their latest versions: @ui5/mcp-server 0.2.11 → 0.2.17, hana-mcp-server 0.3.1 → 0.3.4, @sap-ux/fiori-mcp-server 1.4.0 → 1.11.7, and @mariodefe/sap-datasphere-mcp 1.2.1 → 1.5.2 (targets the first non-deprecated release since 1.3.0, lean tool profile default documented). Refreshed sap-mcp-inventory.json and sap-mcp-security.md evidence so all 5 MCP packages now report status=current.

Added

  • Wired CI audit jobs for audit-mcp-freshness.mjs (--enforce flag) and smoke-local-fixtures.mjs as PR/push-triggered workflows (mcp-freshness.yml, local-fixture-smoke.yml), and restored the CodeQL weekly schedule.

Fixed

  • Removed dead scripts that were strict subsets of other tooling: report-plugin-inventory.sh, skills-lock.json, and audit-skills.mjs (covered by audit-effectiveness.mjs).

[2.4.0] - 2026-07-17

Added

  • sap-bw-query: read-only InfoProvider metadata read (bw_describe_provider now returns characteristics with dimension groups, key figures, and dimensions) through capability-gated BWMT 1.27.36 connectivity services, with a javap-verified API map reference (references/bwmt-api-map.md).
  • sap-bw-query: metadata-verified specification validation — bw_resolve_and_validate_spec accepts a connection alias, verifies every referenced characteristic/key figure against the provider, and reports blocking gaps plus readyForDraft/metadataChecked.
  • sap-bw-query: QuerySpec v1 revision 1.1 with concrete schemas for restricted/calculated/formula key figures, structures, hierarchies, conditions, exceptions, display settings (zero suppression, sign presentation, result position), and enriched variables, including cross-reference validation.
  • sap-bw-query: bw_populate_query_editor builds the confirmed draft inside the wizard-created, still-unsaved query editor (axes, key-figure structures, restricted/formula members, filters, variable bindings, conditions, exceptions, zero suppression) as one undoable editing-domain command with a per-element apply report; saving stays a human action.
  • sap-bw-query: deep query model read — bw_read_query_model (capability-gated readQueryModel bridge call plus QueryModelReader) serializes an open query's complete EMF model to a stable read-only JSON contract with a serializationIssues list, and bw_populate_query_editor now runs it automatically after a successful population to merge a non-fatal verification summary (VERIFIED/DIVERGED/UNAVAILABLE with per-element checks) that compares the confirmed spec against the live model; the human-save boundary is unchanged (saved: false).
  • sap-bw-query: shared best-practices rule engine (mcp/src/query-rules.mjs, catalog BWQ001–BWQ012) that reviews both draft specs and open queries against BW query-design rules — exposed as the read-only bw_review_query tool (reusing the readQueryModel deep read), an additive bestPractices array on bw_resolve_and_validate_spec, the references/query-design-rules.md catalog, and the /bw-query-review and /bw-query-explain commands; everything read-only with the safety contract unchanged.
  • sap-bw-query: metadata-aware optimization proposals (VARIABLE_REUSE, RESTRICTED_MEASURE_CONSOLIDATION, HIERARCHY_DISPLAY_LEVEL, AGGREGATION_CONFIRMATION, FREE_AXIS_CANDIDATE) tagged as semantics-preserving or semantics-changing.
  • sap-bw-query: live-validation runbook (docs/project/sap-bw-query-live-validation-runbook.md); execution remains pending and the verification ledger keeps live BW validation marked pending.
  • sap-bw-query: round-trip smoke that builds an in-memory query with QueryModelBuilder, reads it back with QueryModelReader, and asserts model round-trip integrity fully offline (no workbench, no save). It runs as plain Java (StandaloneSmoke) against a curated BWMT model classpath in CI after every bundle build — reliable and fast, unlike launching the full Eclipse product headless; an in-OSGi SmokeApplication variant also exists. Plus a QuerySpec template library (references/query-templates/ — plan/actual variance, YTD comparison, top-N, reconciliation totals — with references/query-templates.md) whose starting-point specs validate with no warning-severity design-rule finding.
  • sap-bw-query: the bundle builder (bundle/Build-BwStudio.ps1) now publishes the finished release ZIP and manifest (plus signature files when signed) to the local user Desktop (%USERPROFILE%\Desktop) by default after a successful build, then opens File Explorer on the output directory with the artifact selected; it refuses to auto-publish into a OneDrive-managed Desktop (warning and leaving the artifacts in the output directory), an explicit -PublishDirectory always wins, -SkipPublish disables publishing, -SkipExplorer (and automatic CI/non-interactive detection) disables the Explorer launch, and any publish or Explorer failure only warns without failing the build.

Changed

  • Bumped the coordinated repository, marketplace, plugin manifest, and skill metadata versions to 2.4.0.
  • sap-bw-query internal bundle/plugin/MCP version 0.1.0 → 0.3.0 (0.2.0 carried the metadata read and editor population; 0.3.0 adds the deep query model read, review surfaces, and the headless smoke application; a rebuild and redeploy is required to pick up each new Eclipse plug-in and MCP server).
  • sap-bw-query: editor scans now run on the SWT display thread, fixing open-editor inspection (bw_list_queries, bw_read_query, bw_describe_provider openQueries) that previously always returned empty results from the bridge thread.
  • sap-bw-query: a successful deploy now writes two launch shortcuts to the user's visible desktop (which follows OneDrive/Known-Folder-Move redirection) so the studio — installed under the per-user, no-admin %LOCALAPPDATA%\BWAutomationStudio — is discoverable: SAP BW Automation Studio (kein Passwortspeicher).lnk (eclipse.exe -noPwdStore, recommended) and SAP BW Automation Studio (mit Passwortspeicher).lnk (no flag, so Eclipse's own secure store may remember credentials — manual login only; the automation/MCP surface still never handles passwords). Created via COM WScript.Shell (no admin, no deletion), refreshed each deploy, best-effort, and skipped in CI / when BW_STUDIO_NO_SHORTCUT is set.
  • sap-bw-query: the builder additionally drops a small .lnk to the (local) published bundle ZIP on the visible desktop, so the finished ~1 GB build is visible and one-click accessible while only a few KB sync to OneDrive instead of the whole archive.
  • sap-bw-query: the deployer (scripts/BwStudio.ps1) now handles a same-version redeploy with changed content instead of silently no-op'ing. It records the physical install folder as installDir on the activation record and resolves the runtime from it; an unchanged artifact is reused idempotently, while a same-version build with different content installs into a content-addressed sibling folder (versions/<version>+<sha>) and is activated, leaving every prior folder untouched (still strictly append-only, no deletion). Start-BwMcp.ps1, launch, status (activeInstallDir), and rollback all resolve via installDir with a fallback for older records.

Fixed

  • sap-bw-query: QueryModelBuilder resolved BWMT EEnums via getByName(javaConstantName), but the model enums expose CamelCase names/literals (e.g. EQUAL→Equal, TOP_N→TopN, BAD1→Bad1, FOR_ALL_VALUES→forAllValues), so every lookup returned null — comparison/condition/exception operators, alert levels, sign presentation, and zero-suppression mode would all have failed to apply on a live populate. Replaced all six sites with a normalization-based resolver that matches the Java constant name, EMF name, or literal; verified end-to-end by the round-trip smoke on real EMF objects (found via the live smoke; offline JS tests never exercised the real enums).
  • sap-bw-query: the Activator no longer starts the interactive named-pipe bridge (and no longer subclasses AbstractUIPlugin) when -Dbw.automation.headless=true, so headless entry points do not drag in the workbench/bridge machinery.

[2.3.2] - 2026-06-20

Changed

  • Bumped the coordinated repository, marketplace, plugin manifest, and skill metadata versions to 2.3.2.

Fixed

  • Removed redundant manifest-level hook path declarations and switched hook activation to Node-based validator entrypoints to avoid Claude Code hook loader crashes while keeping hook sidecars portable across supported operating systems.

[2.3.1] - 2026-06-19

Added

  • Added sap-sac-test-automation for SAP Analytics Cloud browser and Playwright automation, including dashboard profile/scenario guidance, governance and testability notes, failure-triage artifacts, Edge/CDP patterns, Chrome DevTools MCP coverage, and /sac-test-onboard.
  • Added sap-rpt1 for SAP-RPT-1-OSS FI/CO local tabular prediction workflows, including prepare/predict commands, sample FI/CO datasets, leakage checks, governance guidance, enterprise portability notes, and source-review evidence.
  • Added CAP CDS LSP launcher support and exposed the CAP LSP configuration through generated plugin and marketplace metadata.
  • Added SAC custom widget design/runtime guidance, AI-assisted composite generation notes, CSS/styling compliance coverage, browser runtime references, and design-runtime templates.
  • Added an SAP Integration Suite HTTPS-to-SFTP iFlow package authoring reference and reusable template package.
  • Added Oracle shared review tooling and multi-harness portability documentation for using the same SAP skills across supported AI coding assistants.

Changed

  • Bumped the coordinated repository, marketplace, plugin manifest, and skill metadata versions to 2.3.1.
  • Updated README and current-state marketplace documentation for the 37-plugin inventory: 64 commands, 31 agents, 8 hook-enabled plugins, 6 MCP-enabled plugins, and 1 LSP-enabled plugin.
  • Strengthened validation coverage for manifest drift, MCP security and environment contracts, packaging hygiene, public claims, Oracle browser safety, reference provenance, templates, command/agent contracts, hook contracts, and skill quality.
  • Improved multi-harness wording and evidence language so Claude-specific sidecars are described as portable guidance where clients do not support them natively.

Fixed

  • Removed outdated website planning documents and stale marketplace backup artifacts from the packaged repository state.
  • Addressed SAP-RPT-1 follow-up review items, including third-party attribution cleanup.

[2.3.0] - 2026-06-14

Added

  • Added 18 targeted slash commands for ABAP Cloud review, ABAP CDS model checks, AI Core deployment checks, Cloud SDK AI chat templates, BTP destination/logging/service-manager/transport diagnostics, dependency upgrade planning, Fiori generation/preview checks, HANA connection/object inspection, UI5 CLI build/troubleshooting, and UI5 Linter check/fix planning.
  • Added 2 specialist agents: fiori-app-advisor and hana-database-advisor.
  • Added 14 third-pass slash commands for API style, BTP architecture/readiness, Work Zone, BAS, CIAS, Cloud Identity Services, Integration Suite, ISA export, Job Scheduling, MDI, Data Intelligence, and HANA ML planning checks.
  • Added 4 third-pass advisory agents: api-style-reviewer, btp-platform-advisor, identity-security-advisor, and integration-flow-advisor.
  • Added the sap-dependency-security hook profile to flag unpinned executable package specs, @latest, and credential-like literals in dependency and MCP configuration files.
  • Added npm run audit:skills as a read-only per-plugin capability and verification-status report.
  • Added npm run audit:effectiveness as a read-only report for trigger quality, README drift, oversized references, command contracts, capability inventory, and next recommended artifacts.
  • Added docs/project/plugin-skills-third-pass-audit-2026-06-14.md documenting per-plugin findings, docs-only evidence, and unresolved production-only checks.

Changed

  • Renamed dependency-upgrade to sap-dependency-security as the canonical SAP dependency security and MCP executable trust policy plugin.
  • Added SAP/MCP dependency policy coverage for MCP server pins, SAP Node tooling, Java/Maven, Gradle, Python, containers, BTP/CF/mbt, and ABAP/gCTS review workflows.
  • Exposed plugin sidecar capabilities in generated public interfaces: plugin manifests and marketplace entries now carry hooks and mcpServers when present.
  • Standardized command and agent metadata, including explicit tool scopes, input hints, trigger-rich descriptions, MCP fallback guidance, and tenant/write safety constraints.
  • Standardized all slash commands with an Output Contract and read-only/non-mutating defaults unless explicitly generation or fix-planning commands.
  • Added compact capability indexes to all plugin-local README files and routed oversized references through SKILL.md search guidance.
  • Strengthened validation for packaged artifacts, hook/MCP manifest consistency, command/agent quality, nested manifests, dependency-security hook cases, README capability drift, oversized-reference routing, command output contracts, and mutating-default safeguards.
  • Bumped the marketplace and all generated plugin manifests to version 2.3.0.
  • Documented final third-pass inventory: 35 plugins, 61 commands, 30 agents, 8 hook-enabled plugins, 6 MCP configs, and 1 LSP config.
  • Removed the local .agents mirror check from active validation; .agents/skills/dependency-upgrade remains a local mirror only.

Fixed

  • Addressed PR #82 review findings by pinning the Trivy action template to aquasecurity/trivy-action@v0.36.0 and making its checklist valid YAML comments.
  • Updated the Python security template so uv projects use uv audit and only run pip-audit after exporting requirements.txt.
  • Corrected stale SAPUI5 and Datasphere MCP version references to match the approved exact pins.
  • Removed generated hook bytecode artifacts from packaged plugins and made validation fail on future __pycache__/, *.pyc, backup/temp, and .DS_Store artifacts.

[2.2.3] - 2026-06-12

Added

  • sap-cloud-sdk-ai-python: New plugin at plugins/sap-cloud-sdk-ai-python with comprehensive guides and orchestration support
    • 5 reference docs: getting-started-auth, langchain-guide, native-clients-guide, orchestration-guide, troubleshooting
    • 385-line SKILL.md covering SAP Cloud SDK for AI Python client
  • sap-btp-cloud-identity-services: New plugin at plugins/sap-btp-cloud-identity-services with core documentation references
    • 4 reference docs: app-integration-patterns, authorization-management, cloud-foundry-security, xsuaa-integration-guide
    • 165-line SKILL.md covering authentication, authorization, and identity patterns on BTP
  • sap-ai-core: Expanded AI Launchpad guide, generative AI hub, model providers, and orchestration modules references
  • sap-btp-best-practices: Added ai-development-best-practices.md reference (155 lines)
  • sap-cap-capire: Added comprehensive patterns for CAP integration with SAP Cloud SDK for AI including asynchronous processing guidelines
  • sap-cloud-sdk-ai: Added 59 lines of new patterns for AI integration
  • sap-btp-developer-guide: Added AI development best practices references

Changed

  • sap-sac-scripting: Refreshed to QRC2 2026 (SAC 2026.8, OSE API v2025.20)
    • Skill version bumped 3.0.1 → 3.1.0
    • New references/whats-new-qrc2-2026.md covering Q2 2026 scripting changes
    • All 8 OSE API reference files updated from v2025.14 to v2025.20
    • Added 13 missing OSE API classes (DataUpload*, Variance*, Compass, Math, DataLabelType)
    • Reference count: 63 → 65
  • sap-sac-custom-widget: Refreshed to QRC2 2026
    • Documented missing schema properties in json-schema-reference.md
    • Expanded advanced-topics.md with 49 new lines
    • Updated README and SKILL.md
  • sap-sac-planning: Refreshed to QRC2 2026 (SAC 2026.8)
  • README.md: Restructured with universal installation via vercel-labs/skills, compacted plugin tables, added table of contents

[2.2.2] - 2026-05-30

Added

  • Added dependency-upgrade plugin at plugins/dependency-upgrade and hooked it into .claude-plugin/marketplace.json generation.
  • Added dependency hardening references to high-confidence dependency-using SKILLs in:
    • sap-cloud-sdk-ai
    • sap-hana-cli
    • sap-fiori-tools
    • sap-cap-capire
    • sap-datasphere
    • sap-btp-job-scheduling
    • sapui5-linter
    • sapui5
    • sap-hana-ml
    • sap-sac-scripting
    • sapui5-cli
  • Added dependency security workflow:
    • .github/workflows/dependency-security-checks.yml
    • deterministic policy enforcement for dependency-manageable skill and lockfile/PM changes
  • Replaced plugins/dependency-upgrade generator placeholder implementation with functional script:
    • .agents/dependency-upgrade/scripts/generate-dependency-upgrades.sh
  • Added dependency hardening reference notes in .agents workflow/automation docs

[2.2.1] - 2026-05-30

Fixed

  • Issue #74: Removed blocking UserPromptSubmit hooks from sap-sqlscript and sap-sac-custom-widget that were stopping unrelated prompts (e.g., "hello"). Both skills now rely on non-blocking, command-based validation hooks only.

[2.1.8] - 2026-04-02

Added

  • Frontmatter validation infrastructure ported from claude-skills
    • scripts/validate-frontmatter.sh: comprehensive SKILL.md YAML frontmatter checker
    • .github/workflows/validate-frontmatter.yml: CI workflow (diff on PR, full scan on push)
    • .github/workflows/validate-json-schemas.yml: CI workflow with artifact upload on failure
    • .githooks/pre-push: validates changed SKILL.md files before push
    • Enhanced .githooks/pre-commit with --spec=draft7 consistency
    • Split quality-checks.yml into focused workflows

Fixed

  • sapui5, sapui5-linter: Fixed YAML frontmatter parsing errors preventing skill installation (#63)
  • sap-abap-cds, sap-btp-cloud-platform: Fixed description overflow preventing skill installation (#65)
  • sap-btp-connectivity, sap-btp-integration-suite, sap-datasphere, sap-sac-custom-widget: Fixed description overflow (>1024 chars)
  • sapui5-linter: Fixed unclosed quote on ui5_linter_version metadata field
  • All 32 skills now pass frontmatter validation

[2.1.7] - 2026-03-14

Changed

  • Hook validators hardening (sap-cap-capire, sap-datasphere, sap-sac-custom-widget, sap-sac-planning, sap-sac-scripting, sap-sqlscript, sapui5): Four CodeRabbit-identified fixes applied to all 14 validator files (validator.mjs + validator.py per plugin):
    • Added stdin error event handler in readStdin() to prevent Promise hang on I/O errors
    • Security risk messages now enumerate all detected risks with explanations, not just the first
    • Function constructor check upgraded to whitespace-tolerant regex (new\s+function\s*\()
    • WHERE clause detection upgraded to word-boundary regex (\bwhere\b) to catch newline-separated SQL

[2.1.6] - 2026-02-26

Added

  • sap-sqlscript: Added /sqlscript-setup command for automated development environment setup
    • New scripts/setup-vscode.sh detects VS Code, installs @sap/hana-sqlscript-lsp VSIX, checks HANA env vars
    • Command parses structured script output, writes/merges .vscode/settings.json, prints summary table
    • Hybrid approach: shell for deterministic steps, LLM for environment summary and settings authoring

Changed

  • Hooks hardening: Replaced broad prompt-based Write|Edit hooks with deterministic command hooks in 7 plugins
    • Plugins: sap-cap-capire, sapui5, sap-datasphere, sap-sac-planning, sap-sac-scripting, sap-sac-custom-widget, sap-sqlscript
    • Added runtime fallback per plugin hook: Node.js primary, Python fallback, fail-open allow mode
    • Enforced structured JSON hook output and security-focused blocking (PreToolUse only); PostToolUse remains advisory
    • Goal: prevent unrelated file edits from stopping continuation in mixed-plugin workflows (issue #60)

[2.1.5] - 2026-02-26

Added

  • sap-sac-scripting: Added MCP integration via community sap_analytics_cloud_mcp
    • New .mcp.json connecting to SAP Analytics Cloud REST API MCP server
    • 90 tools across 11 service areas: Content, Data Export, Data Import, Multi Actions, Calendar, Content Transport, User Management, Monitoring, Schedule & Publication, Translation, Smart Query
    • OAuth 2.0 Client Credentials authentication
    • Requires local clone and build from https://github.com/secondsky/sap_analytics_cloud_mcp
    • Set SAC_MCP_PATH, SAC_BASE_URL, SAC_TOKEN_URL, SAC_CLIENT_ID, SAC_CLIENT_SECRET

[2.1.4] - 2026-02-26

Added

  • sap-fiori-tools: Added MCP integration via @sap-ux/fiori-mcp-server
    • New .mcp.json connecting to SAP's official Fiori MCP server
    • 5 tools: search_docs, list_fiori_apps, list_functionalities, get_functionality_details, execute_functionality
    • Semantic search across Fiori Elements, Annotations, UI5, and Fiori tools documentation
    • AI-assisted programmatic Fiori app generation and modification
    • No authentication required; requires Node.js 20+

[2.1.3] - 2026-02-25

Fixed

  • Reserved words validation: Removed blocked words (official, anthropic, claude) from marketplace and plugin descriptions (fixes #56)
    • marketplace.json: Updated description to avoid CLI installation errors
    • sap-ai-core: Removed model name from description
    • sap-api-style: Changed "official SAP API Style Guide" to "the SAP API Style Guide"
    • sap-cloud-sdk-ai: Removed model name from description

Added

  • validate-reserved-words.sh: New validation script to prevent blocked words in name/description fields
  • GitHub Actions: Added reserved words check to quality-checks.yml

Changed

  • CLAUDE.md: Added Reserved Words Policy documentation

[2.1.2] - 2026-02-06

Changed

Documentation Clarity

  • README.md: Updated terminology to clarify distinction between plugins (installable units) and skills (AI capabilities)

    • Changed badge from "Skills" to "Plugins" to reflect what users install
    • Added subtitle: "32 Claude Code plugins for SAP development"
    • Updated installation instructions to use "plugins" terminology
    • Clarified that "plugins provide skills that auto-activate"

  • Plugin Tables: Enhanced all 5 category tables with component visibility

    • Added Component Legend explaining Commands, Agents, and MCP features
    • Added 3 new columns (Commands | Agents | MCP) to show what each plugin offers
    • Component counts: 7 plugins with commands (28 total), 7 with agents (24 total), 4 with MCP integration
    • Clear indicators: Numbers for commands/agents, checkmark (✓) for MCP, em-dash (—) for none

  • Content Updates Throughout:

    • "Building New Skills" → "Building New Plugins" section title
    • Repository Structure: Updated to accurate component counts
    • Contributing section: Clarified plugin vs. skill terminology
    • Success Metrics: Updated "skill creation/testing" to "plugin" terminology
    • 15+ terminology updates across all sections for consistency

Why This Update

This release clarifies the conceptual model for users:

  • Plugins are what you install via /plugin install
  • Skills are the AI capabilities that activate automatically when you work
  • Both terms are correct in their respective contexts

Component visibility helps users:

  • Quickly identify which plugins offer advanced features (commands, agents, MCP)
  • Understand the scope of each plugin's capabilities before installation
  • Make informed decisions based on needs (basic skills vs. advanced automation)

[2.1.1] - 2026-02-06

Changed

  • Updated author field from "SAP Skills Maintainers" to "E.J." across all 64 plugin.json files
  • Added Claude Code-specific clarification to README.md explaining that skills follow Claude Code plugin patterns
  • Updated contributor guide to reflect manual review process
  • Removed skill-review automation references from all documentation
  • Added deprecation notices to quality-assurance.md and skills-review-progress.md
  • Updated workflow-checklist.md with manual quality review guidelines

Removed

  • skill-review plugin - Automated quality assurance plugin removed (transitioned to manual review process)
  • Reduced total skills: 33 → 32
  • Removed from Tooling & Development category (4 → 3 skills)

[2.1.0] - 2025-12-27

Major Changes

Plugin Structure Migration

  • BREAKING: Migrated from flat skills/ directory to plugin-based architecture
  • Repository structure changed from skills/[name]/ to plugins/[name]/skills/[name]/
  • All 33 skills successfully migrated and validated
  • Generated 91 JSON manifest files (plugin.json at plugin and skill levels)
  • Dual-level plugin architecture: plugin-level and skill-level manifests

Added

Advanced Plugin Features

  • Agent Support: 5 plugins now include specialized agents for autonomous task execution

    • sap-datasphere: 3 agents (datasphere-admin-helper, datasphere-integration-advisor, datasphere-modeler)
    • sap-sac-custom-widget: Custom widget development agents
    • sap-sac-planning: Planning application agents
    • sap-sac-scripting: Scripting automation agents
    • sap-sqlscript: SQL optimization and performance agents

  • Command Support: 5 plugins with slash commands for quick access

    • sap-datasphere: /datasphere-cli, /datasphere-connection-guide, /datasphere-space-template, /datasphere-view-template
    • Additional command support for SAC and SQLScript plugins
    • Instant templates and guides via command invocation

  • Hook Support: 5 plugins with validation and automation hooks

    • Pre-tool-use validation hooks
    • Post-tool-use automation hooks
    • Quality gate enforcement
    • Session lifecycle management

  • MCP Server Integration: 3 plugins with Model Context Protocol support

    • sap-cap-capire: CAP MCP server (@cap-js/mcp-server)

      • Live access to compiled CDS models
      • Semantic search for entities, services, actions, relationships
      • CAP documentation query capabilities
      • Zero configuration required

    • sap-datasphere: Datasphere MCP server (@mariodefe/sap-datasphere-mcp)

      • Direct integration with SAP Datasphere API
      • Requires authentication via environment variables
      • Access to spaces, views, models, and connections

    • sapui5: UI5 Tooling MCP server (@ui5/mcp-server)

      • UI5 project tooling integration
      • Version-aware assistance (UI5 1.120.0+)
      • Project structure awareness

Enhanced Documentation

  • 22 new reference files across plugins providing deep-dive documentation

    • sap-datasphere (5 files):

      • best-practices-patterns.md - Design patterns and architectural guidance
      • catalog-governance.md - Data catalog and governance practices
      • cli-commands.md - Complete CLI reference
      • data-products-marketplace.md - Data product management
      • whats-new-2025.md - Latest features and updates

    • sap-sac-planning (4 files):

      • bpc-live-connection.md - BPC integration patterns
      • data-action-tracing.md - Debugging and tracing
      • seamless-planning-datasphere.md - Datasphere integration
      • value-driver-trees.md - Advanced planning scenarios

    • sap-sac-scripting (4 files):

      • chart-variance-apis.md - Chart customization APIs
      • compass-seamless-planning.md - Compass planning integration
      • data-actions-enhancements.md - Data action best practices
      • whats-new-q4-2025.md - Latest scripting features

    • sap-sqlscript (2 files):

      • glossary.md - SQLScript terminology reference
      • skill-reference-guide.md - Complete skill documentation

    • Additional template directories for rapid scaffolding

Marketplace Enhancements

  • Cross-references: 13 plugins with comprehensive cross-skill references

    • sap-abap ↔ sap-abap-cds, sap-cap-capire, sap-fiori-tools, sap-btp-cloud-platform, sap-api-style
    • sap-abap-cds ↔ sap-abap, sap-btp-cloud-platform, sap-fiori-tools, sap-cap-capire, sap-api-style
    • sap-btp-cloud-platform ↔ sap-btp-best-practices, sap-cap-capire, sap-fiori-tools, sap-ai-core, sap-abap
    • sap-btp-best-practices ↔ sap-btp-cloud-platform, sap-btp-connectivity, sap-btp-service-manager
    • sap-fiori-tools ↔ sapui5, sap-cap-capire, sap-abap-cds, sap-btp-cloud-platform
    • sapui5 ↔ sap-fiori-tools, sap-cap-capire, sap-btp-cloud-platform
    • sap-cap-capire ↔ sap-fiori-tools, sapui5, sap-btp-cloud-platform, sap-hana-cli, sap-abap
    • sap-hana-cli ↔ sap-cap-capire, sap-btp-cloud-platform, sap-abap-cds, sap-datasphere
    • sap-ai-core ↔ sap-btp-cloud-platform, sap-cap-capire, sap-cloud-sdk-ai
    • sap-api-style ↔ sap-cap-capire, sap-fiori-tools, sap-abap, sapui5
    • sap-btp-connectivity ↔ sap-btp-cloud-platform, sap-btp-best-practices, sap-cap-capire
    • sap-btp-service-manager ↔ sap-btp-cloud-platform, sap-btp-best-practices, sap-btp-connectivity
    • sap-btp-developer-guide ↔ sap-btp-cloud-platform, sap-btp-best-practices, sap-cap-capire
    • Full cross-reference network documented in CLAUDE.md

  • Enhanced Metadata: All skills include comprehensive keywords for improved discovery

  • Category Organization: Skills organized into 5 categories (Tooling, BTP, UI, Data, Core)

Changed

Infrastructure Updates

  • Scripts: Updated generation and synchronization scripts

    • generate-plugin-manifests.sh: Generates plugin.json from SKILL.md YAML frontmatter
    • sync-plugins.sh: Unified version sync, manifest generation, and marketplace update
    • generate-marketplace.sh: Creates marketplace.json from all plugin.json files
    • All scripts support dry-run mode for safety

  • Version Alignment: All 33 skills synchronized to version 2.1.0 or higher

  • Directory Structure: Standardized plugin structure across all skills

  • Manifest Generation: Automated dual-level manifest creation

Documentation Structure

  • Plugin Paths: All documentation updated to reflect plugins/[name]/skills/[name]/ structure
  • README Updates: Skill-level README files enhanced with auto-discovery keywords
  • SKILL.md Updates: Aligned with latest Anthropic skills specification
  • Cross-Linking: Improved internal documentation cross-references

Fixed

  • Quality Checks: Removed failing quality checks workflow
  • Path References: Fixed QUICK_WORKFLOW.md link paths in documentation
  • Backup Cleanup: Removed SKILL.md.backup files after migration
  • Manifest Consistency: Ensured all plugin.json files have consistent structure
  • Marketplace Paths: Updated all source paths in marketplace.json

Documentation

  • CLAUDE.md: Comprehensive project guidelines with critical review instructions

    • Manual review requirement (no automation scripts for refactoring)
    • Skill-review skill usage for quality audits
    • Current status documentation (33 production-ready skills)
    • Marketplace cross-references documentation
    • Quality standards and verification checklist

  • migra.md: Complete migration guide for plugin structure

    • Problem diagnosis and solution
    • Correct plugin structure examples
    • Migration steps and validation

  • README.md: Updated to reflect new plugin architecture

    • Installation instructions
    • Skill auto-trigger examples
    • 33 skills organized by category
    • Plugin structure diagram
    • Building new skills guide
    • Documentation links

  • CHANGELOG.md: Introduced this changelog following Keep-a-Changelog format

[2.0.0] - Previous Versions

Earlier versions used a flat skills/ directory structure. Migration to plugin architecture completed in v2.1.0.

Historical Structure:

  • Skills located at skills/[name]/
  • Single-level manifest files
  • Basic marketplace integration

Migration Notes:

  • All v2.0.0 skills successfully migrated to v2.1.0 plugin structure
  • No feature regressions during migration
  • Enhanced capabilities added (agents, commands, hooks)

Upgrade Guide

From v2.0.0 to v2.1.0

If you have local modifications or custom skills:

  1. Backup your work:

    git stash

  2. Pull the latest changes:

    git checkout main
git pull origin main

  3. Update local custom skills to new structure:

    # Old structure: skills/my-skill/
# New structure: plugins/my-skill/skills/my-skill/

mkdir -p plugins/my-skill/skills/
mv skills/my-skill plugins/my-skill/skills/my-skill

  4. Generate manifests:

    ./scripts/sync-plugins.sh

  5. Verify:

    jq '.plugins | length' .claude-plugin/marketplace.json
# Should show total plugin count including your custom skills

Links

Maintained by: E.J.