Skip to content

SkillMeat v0.18.0 Release Notes

Release Date: 2026-03-29

Added

  • Backstage Unified Telemetry Views (P0-P1): Entity Intelligence Tab and Platform Health Dashboard with enterprise-wide KPI visibility.
  • EntitySkillMeatIntelligenceContent: New unified Intelligence Tab combining deployment context (Context, BOM, Drift cards) with telemetry cards (Workflow Effectiveness, Agentic Metrics) in responsive 2-column grid layout. Telemetry columns gated by backstage_analytics_widgets feature flag; core cards always visible.
  • PlatformHealthDashboardPage: New full-page dashboard with KPI header row (Total Projects, Avg Effectiveness Score, Cost MTD, Drift Alerts), enhanced Global ROI Dashboard with row-click entity navigation to Intelligence Tab, catalog health SVG donut chart, and top drifted projects panel. KPI header row feature-flagged.
  • Analytics Summary Endpoint: GET /api/v1/analytics/project-summary/{project_id} provides project-level aggregated metrics (effectiveness scores, cost data, drift metrics) for dashboard population.

  • Feature Flag Routing: Platform engineers control telemetry widget visibility via backstage_analytics_widgets feature flag without code changes.

  • Artifact Detection & Import Expansion (Phases 1-7): Complete expansion of artifact detection system from 7 to 18 artifact types with cross-platform support and tier-based organization.

  • Type Expansion: Extended artifact detection from Skills, Commands, Agents, Hooks, MCP servers, and Workflows to 18 types across 4 tiers
    • Tier 0 (Context Entities): PROJECT_CONFIG, SPEC_FILE, RULE_FILE, CONTEXT_FILE, PROGRESS_TEMPLATE
    • Tier 1 (Composites): CONTEXT_MODULE, TEMPLATE, GROUP
    • Tier 2 (Advanced): DEPLOYMENT_SET, CONTEXT_PACK
    • Tier 3 (Enhanced): BUNDLE
  • Core System: Enhanced ArtifactSignature dataclass with container_names, manifest detection, nesting support, and confidence scoring (container +30, manifest +40, file pattern +30; threshold 80 for strict mode)
  • Context Entity Visibility: Rules, Specs, Context Files, Memory files, Config Files, and Progress Templates now discoverable as first-class artifacts in local projects; stored in platform-specific directories (.claude/, .cursor/, .codex/, .gemini/)
  • Platform Variation Support: Detection works across Claude Code, Cursor, Codex, and Gemini platform layouts with automatic container name normalization (e.g., context-modules/modules/context_modules all map to CONTEXT_MODULE)
  • Heuristic Detection: Smart detection for Tier 1-3 types via directory structure and manifest files; includes file-pattern-only detection (e.g., .template.md for single-file templates)
  • Strict Mode Enforcement: GROUP artifacts require manifest files (confidence floor 75+); prevents false positives on empty directories
  • Universal Frontmatter Envelope: Optional skillmeat_type frontmatter field for explicit type declaration per ADR-009 (manual override for detection edge cases)
  • TypeScript Type Sync: Frontend ArtifactType union now synchronized with Python enum (18 types) with exhaustiveness checks in Record maps
  • Tier-Based Grouping: Frontend supports grouping artifacts by their tier classification (context entities, composites, advanced, enhanced) for organized browsing
  • Database Schema: Alembic migration adds artifact_tier column to track tier membership; tier map stored in SQLAlchemy ORM

  • API Contract: OpenAPI schema updated with 18 artifact types across all responses; ArtifactMetadataResponse includes tier information

  • Bob Platform Support: Added IBM Bob as a new built-in platform profile, identical to Claude Code but using .bob/ project directory. Includes full backend support (Platform enum, platform defaults, path resolver, deployment fallback), frontend constants and display names, and SDK type update.

  • Custom Profile Documentation: Added comprehensive guide for creating custom deployment profiles via Web UI, CLI/API, and global Settings, including practical curl examples.

  • Tiered Card System: New unified ArtifactCard component with 5 tiers (xs-minified, minified, compact, standard, expanded) replacing 3 legacy card components.

  • Tier-based zone activation: Header, Content, Relationship, Status, Action zones render conditionally based on tier
  • Type-coded left border accent (border-l-4) for visual artifact type differentiation across all tiers
  • Colored tag badges with DB-first color resolution and hash-based fallback, capped per tier (3-5 tags)
  • Group badges showing artifact memberships with tier-appropriate limits (2-4 groups)
  • Standalone ArtifactActions dropdown component for use in list/table row contexts
  • DraggableArtifactCard wrapper with @dnd-kit integration for drag-and-drop support in minified tier
  • TagColorProvider context to prevent n+1 useTags hook calls across card grids
  • Background tints for artifact type identification at compact and larger tiers
  • Composite member icons and platform badges at standard tier

  • Full documentation in components/artifacts/cards/README.md

  • Group View Mode: Browse artifacts organized by groups with drag-and-drop reordering.

  • New 'group' view mode in artifact browser (alongside Grid and List)
  • ?view=group and ?group=<id> URL state for persistent view/group selection
  • Drag-and-drop to move artifacts between groups with visual feedback

  • Available in both Library and Operations lenses

  • Enhanced View Layouts: Operations lens grid now renders multi-column layout; enhanced list view with full action set.

  • Operations lens grid: responsive 2 columns at 768px, 3 columns at 1280px

  • Enhanced list view with type-colored icons, tag badges, group badges, and full action menu

  • MkDocs Documentation Site: Unified documentation site using MkDocs Material with integrated search, navigation tabs, and light/dark theme toggle.

  • CLI reference automatically generated via mkdocs-click plugin from Click introspection (always in sync)
  • 57 existing documentation files organized into 14-section navigation hierarchy
  • Consumer documents updated to cross-link auto-generated CLI reference
  • Build validation via mkdocs build --strict, local preview via mkdocs serve

  • Doc workflow specs updated to codify MkDocs-based maintenance pattern

  • SkillBOM (Bill of Materials) & Attestation Framework: Complete implementation (Phases 1-11) with BOM generation, Ed25519 signing/verification, attestation records, RBAC scoping, history tracking, and time-travel restore.

  • 13+ artifact type adapters (Skills, Commands, Agents, MCP servers, Hooks, Composites, Projects, Deployments, Groups, Tags, Settings, Context entities)
  • BOM serialization with atomic file writes and deterministic sorting
  • Ed25519 cryptographic signing and verification with key generation
  • Attestation records with owner/team/enterprise scoping and policy enforcement
  • Artifact history tracking with event categorization and diff payloads
  • Time-travel BOM restore from historical snapshots
  • Pydantic v2 schemas (BomSchema, AttestationSchema, HistoryEventSchema)
  • 8 REST API endpoints (BOM snapshot, generate, verify, attestation CRUD)
  • 6 CLI commands (generate, sign, verify, keygen, restore, install-hook)
  • React components (ProvenanceTab, BomViewer, AttestationBadge, ActivityTimeline)
  • Backstage plugin (SkillBOMCard, scaffolder actions)
  • Feature flags: skillbom_enabled, skillbom_auto_sign, skillbom_history_capture
  • 605 tests (unit, integration, migration, load, feature flags) with >= 80% coverage

  • User guides, API documentation, and phased rollout plan

  • Enterprise Mode Repository Parity: Full dual-mode support (local + enterprise) with 18 new SQLAlchemy ORM models, Alembic migrations, and enterprise-specific repositories for all data layers.

  • EnterpriseDbCollectionArtifactRepository, EnterpriseProjectRepository, EnterpriseDeploymentRepository
  • EnterpriseTagRepository, EnterpriseGroupRepository, EnterpriseSettingsRepository
  • EnterpriseMarketplaceSourceRepository, EnterpriseProjectTemplateRepository
  • Tenant isolation + multi-tenant session management

  • Dual SQLAlchemy mode (1.x for local, 2.x select() for enterprise)

  • Managing READMEs Skill: New skill for best practices in managing AI artifact documentation.

  • 14 items covering documentation structure, examples, metadata, and maintenance

  • Integrated skill-creator review process

  • @skillmeat/content-viewer Package: Extracted reusable content viewing components into a standalone workspace package with adapter-based architecture.

  • Phase 1-3: Initial extraction of FileTree, ContentPane, FrontmatterDisplay, SplitPreview, MarkdownEditor components with adapter abstraction and parity test suite
    • Adapter abstraction (ContentViewerAdapter interface) for backend decoupling
    • ContentViewerProvider context for adapter injection
    • Frontmatter utilities (parse, strip, detect) and README extraction helpers
    • 133 tests (78 parity, 55 accessibility) with zero regressions
    • Consumer documentation with integration examples
    • 875 lines of duplicate code removed from app, replaced with 3-line re-export stubs
  • Phase 4: Additional Modal Viewer Wave extraction (Wave-002)
    • Extracted DiffViewer (~818 LOC) and FilePreviewPane (~279 LOC) into the package
    • Added generic FileDiff type, badge UI primitive, and perf-marks utility to support new components
    • Created parity test suites (165 tests across 7 suites) for DiffViewer and FilePreviewPane
    • Added accessibility audit tests for both components (wave-002.a11y.test.tsx)
    • All wave components pass zero-regression parity validation
  • Phase 5: Stabilization and cleanup
    • Removed 6 legacy re-export stubs (content-pane, file-tree, frontmatter-display, diff-viewer, file-preview-pane, frontmatter.ts)
    • Migrated all 11+ consumer files to import directly from @skillmeat/content-viewer package
    • Updated consumer documentation (README, API.md) with DiffViewer and FilePreviewPane usage examples

  • Bundle size: 85.3KB (within 110KB limit)

  • Scaffold Template Management (v1): End-to-end template creation and rendering with CLI, web UI, and Backstage integration.

  • New skillmeat scaffold CLI command — renders a Bundle's scaffold template into a target project directory with --dry-run support, deployment registration, and byte-identical parity with Backstage scaffold action
  • New skillmeat template CLI command group — list, create, configure, preview subcommands for managing scaffold templates
  • New /templates web UI page — platform engineers can browse, configure, preview, and publish scaffold templates without writing YAML
  • New /api/v1/templates/scaffold/ API endpoints — full CRUD + preview + enable/disable for scaffold templates
  • New GET /api/v1/integrations/idp/templates discovery endpoint — Backstage EntityProvider polls this for dynamic template catalog entries
  • Template versioning — tracks bundle_version_at_last_generation with stale annotation when Bundle updates ahead of template
  • Dynamic Backstage discovery — per-entity retrieval, platform-filter support, generation caching

  • ScaffoldTemplate ORM model with Alembic migration for template metadata persistence

  • Dev Workflow Process Improvements: Comprehensive multi-model task assignment and implementation plan documentation standardization.

  • Model and Effort columns in implementation plan and phase breakdown task tables for multi-model task assignment
  • UI package extraction spec with decision tree, quality gates, and 9-step extraction process
  • Documentation finalization as a standard prescriptive phase in implementation plans
  • Frontmatter lifecycle guidance showing field population timing across SDLC phases
  • Multi-model routing guidance, UI extraction guidance, and doc finalization guidance references for planning skill

  • External model task grouping pattern (batch_0 pre-work) in dev-execution batch delegation

  • Bundle Deploy Command: New skillmeat bundle deploy <bundle> --project <path> command deploys all bundle members (artifacts + context entities) to a target project directory with per-member status reporting and --dry-run support.

  • Scaffold Bundle Routing: skillmeat scaffold --bundle <name> now routes to the bundle deploy pipeline, deploying both artifacts and context entities instead of only template-rendered artifacts.

  • Project-Aware Init: skillmeat init now creates .skillmeat/manifest.toml in the current directory with project name and UUID when run outside the SkillMeat home directory. Registers project via API (best-effort). Re-running prints "Project already initialized".

  • Embedding Setup Warning: skillmeat match now prints an actionable one-time warning when embedding service is unavailable: Embedding service not configured. Run: skillmeat config set embedding-api-key <key>.

  • Polymorphic Bundle Membership: Bundles can now contain both artifacts and context entities as members via a member_type discriminator on bundle_memberships.

  • skillmeat bundle add-member <bundle> context:<name> adds context entities to bundles
  • skillmeat bundle show displays context entity members under a distinct section

  • BundleResolver.resolve() returns BundleResolutionResult with both artifacts and context_entities lists

  • Frontmatter Fields Exposure: Expose 7 frontmatter fields (effort, model, allowed-tools, disallowedTools, permissionMode, context, scope) in artifact metadata, API responses, and frontend display.

  • Core: ArtifactMetadata dataclass extended with 7 explicit fields, validators for constrained values (effort: dev/research/implementation/review, model: claude-3-5-sonnet/opus/haiku, etc.), key name mapping (kebab-case/camelCase → snake_case)
  • API: All 7 fields added to ArtifactMetadataResponse schema, OpenAPI spec regenerated with complete field documentation
  • Frontend: Fields displayed in artifact detail view (badges for string values, tags for list values, hidden when null); model + effort shown in artifact list view cards for quick reference

  • Tests: Unit tests for validation and parsing, integration tests for API responses, component tests for field rendering, E2E tests for artifact detail page navigation and field visibility

  • Artifact Version Control v2 (Phases 1-4): Complete version history and deployment tracking system with topology visualization, conflict detection, and automated seeding.

  • Version Topology View: New panel showing version deployment status across all projects for any artifact with tier-based project grouping and sync recommendations
  • Push to Collection: Action button to push project's modified artifact version back to collection, with optimistic-concurrency conflict detection (409 response + diff viewer for merge preview)
  • Sync All Outdated: One-click bulk sync of all outdated project deployments to latest collection version, automatically skipping pinned deployments
  • Version Seeding Coverage: All 8 mutation paths now automatically record version history (CLI add, web import, cache refresh, web deploy, CLI deploy, sync, GitHub update, push to collection)
  • Hydration Banner: Auto-running backfill that seeds version history for pre-existing artifacts on first upgrade, with real-time progress polling UI
  • Outdated Projects Badge: Count badge showing outdated project deployments in artifact detail view, with quick-access to Sync All action
  • Database Schema: Three new tables (artifact_versions, artifact_collection_versions, project_version_deployments) with Alembic migration and automatic backfill task
  • API Endpoints: GET /topology, POST /promote, POST /sync-all, GET /backfill-status for version operations and monitoring

  • React Query Hooks: useVersionTopology, usePromoteVersion, useSyncAllOutdated, useBackfillStatus for frontend version management

  • Collection Consolidation v2: Interactive Features: Full artifact management operations in the web UI with content editing, sync workflows, and status dashboards.

  • Critical Actions: Similar artifacts discovery with scored results and navigation; Sync, Deploy, and Delete buttons in artifact modal header; deployments tab with per-project management
  • Content Editing: Inline file content editing with save/cancel; create and delete files with confirmation; unsaved changes guard preventing data loss
  • Sync & Context Management: Project-mode sync supporting collection and project scope; context entity sync status display with merge conflict resolution workflow; rollback to collection version capability
  • Status Dashboard: Health indicator dashboard in new Status tab; deployment summary with project list; quick actions for Sync All, Deploy, and View History
  • Metadata Operations: Tag management (add/remove) via popover; parameter editing modal; parent composite display in Links tab; copy artifact deep link to clipboard; history rollback from timeline entries

  • Accessibility: Keyboard navigation for similar artifact cards; ARIA attributes on context sync status component

  • Marketplace Source Export/Import: Users can export and import marketplace GitHub sources via portable YAML files with preview and automatic deduplication.

  • Export sources to YAML with optional tags/trust levels and scan settings from the marketplace sources kebab menu
  • Import YAML files via drag-and-drop with preview before committing changes; system automatically skips duplicates and creates missing tags
  • Staggered discovery scans prevent API overload when importing multiple sources
  • Portable skillmeat-sources-export-YYYY-MM-DD.yaml format with schema_version: 1 for forward compatibility

  • New POST /api/v1/marketplace/sources/import endpoint for programmatic imports

  • SAM Telemetry Integration: End-to-end telemetry ingestion, aggregation, and Backstage plugin visibility for AI workflow effectiveness metrics.

  • Telemetry ingestion API: POST /api/v1/analytics/execution-outcomes and /project-rollups accept execution telemetry from CCDash instances with PII rejection and rate limiting.
  • Background rollup engine materializes aggregated metrics into project_agentic_metrics within 5 minutes of ingestion.
  • IDP telemetry endpoints: GET /integrations/idp/telemetry/{project_id} and /telemetry/global surface aggregated AI effectiveness metrics to Backstage.
  • Backstage plugin widgets: EntityWorkflowEffectivenessCard, EntityAgenticMetricsContent tab, and Global ROI Dashboard for enterprise-wide AI workflow effectiveness visibility.

  • Feature flags: sam_telemetry_ingestion and backstage_analytics_widgets gate new functionality until CCDash clients are configured.

  • Version-Centric Deployment View: New "By Version" lens on Deployments sub-tab with grouped deployment visualization and batch sync capabilities.

  • Frontend: Toggle between standard and "By Version" lens on Deployments sub-tab; versions grouped hierarchically with deployment status per project; batch sync action for multiple deployments across versions; enterprise owner scope display for project-scoped items
  • Backend: GET /api/v1/versions/artifacts/{id}/topology-by-version endpoint returns version-grouped deployment topology with enterprise scope enrichment; POST /api/v1/versions/artifacts/{id}/batch-sync endpoint executes bulk sync across selected version deployments

Changed

  • Operations Lens Tab Layout: Added lens toggle to horizontal tab layout for operations-lens tab accessibility.
  • Planning Skill Workflow 2: Enhanced with model assignment, UI extraction evaluation, and documentation evaluation steps.
  • Dev-Execution Phase Execution: Added model routing guidance in Stage 2 and documentation delegation in new Stage 6.
  • Progress Template: Now supports optional assigned_model, model_effort, and model_usage fields (backward-compatible).
  • Deployment Cache Invalidation: Added missing cache refresh operations for deployment mutations.
  • Filter 'discovered' sentinel from deployment artifacts

  • Invalidate related cache entries on deployment changes

  • Enterprise API Wiring: Edition-aware initialization for dual-mode deployment support.

  • Conditional repository injection based on edition string ("local" vs "enterprise")

  • Fallback to LocalAuthProvider for zero-config local development

  • Sync Engine: Now hierarchy-aware with forced rollout support for enterprise artifacts and membership-based team visibility filtering.

  • Artifact Cards & Modals: Displays "Enterprise Managed" badge and lock icon for enterprise-owned items with visual hierarchy distinction.

Fixed

  • Enterprise marketplace sources: Full source configuration (description, tags, scan config, visibility, trust level, and 12 other fields) now persists in enterprise PostgreSQL — previously these were property stubs returning None/False. Added concrete ORM methods (create, delete, list_all, get_by_repo_url) to EnterpriseMarketplaceSourceRepository for full router compatibility.
  • Enterprise artifact tag updates (update_collection_tags) now persist in enterprise mode.
  • Enterprise deployment set member reordering (update_member_position) now functional in enterprise mode.
  • Enterprise project template CRUD (list, count, get, create, update, delete, deploy) now fully implemented instead of stubs.
  • Enterprise artifact history get_by_composite_key now returns None instead of raising NotImplementedError.
  • Enterprise artifact repository now implements all IArtifactRepository methods (resolve_uuid_by_type_name, get_ids_by_uuids, get_collection_description).
  • Modal Stub Gap Remediation: Completed wiring of 5 stub modal tabs to real data and functionality.
  • Wire deployment-set-members-tab to real data via useDeploymentSetMembers hook
  • Replace provenance tab "Activity history coming soon" toast with inline activity view
  • Filter context-entities-tab by module_id for module-scoped entity display
  • Verify parameter save in metadata-tab routes through useEditArtifactParameters hook
  • Enterprise Repository Coverage: Fixed missing abstract method stubs in EnterpriseDbCollectionArtifactRepository.
  • Alembic Migration Linearization: Resolved multiple Alembic heads by establishing correct dependency chain for enterprise schema migrations.
  • Enterprise Collection Defaults: Aligned default collection name between local and enterprise modes.
  • Deployment Sentinel Filtering: Fixed discovered artifacts being incorrectly included in deployment lists.
  • Context Add Auto-Frontmatter: skillmeat context add now auto-generates minimal YAML frontmatter for files missing it, instead of rejecting them. Use --strict to enforce frontmatter presence.
  • Context Deploy Content Fetch: skillmeat context deploy now fetches full entity content before writing to disk, fixing empty file output.
  • CLI/API Validator Harmonization: CLI and API now share the same validation function for context entities, ensuring consistent permissive-by-default behavior.

Security

  • RBAC Enforcement: Hardened with enterprise scope isolation preventing cross-tenant data leakage and membership-aware query filters.
  • Enterprise Admin Dependency: New require_enterprise_admin dependency for enterprise-only endpoints with explicit authorization checks.
  • Destructive Sync Gate: Sync operations on enterprise artifacts require explicit user confirmation before executing rollout or override cascades.

Performance

  • Project Discovery: Optimized from 91.35s to 0.57-0.74s (123x improvement)
  • Root cause: 20 separate rglob operations traversing entire directory trees including node_modules, .git, venv, etc.
  • Solution: Single os.walk with in-place directory pruning + asyncio parallelization of cache building
  • Impact: Typical 7-project workspace now loads in less than 1 second instead of over 90 seconds

Previous Releases

For historical release notes, see git tags and branch documentation at https://github.com/miethe/skillmeat.