Skip to main content
EVOKORE// SKILLS / RUNTIME_SUMMARY_v3.0.0
EVOKORE-MCPmarkdownsha 66B22BF9F7A7

RUNTIME_SUMMARY_v3.0.0

documentation

EVOKORE Runtime Summary - v3.0.0

This summary captures the current shipped runtime and operator surface for EVOKORE-MCP at package/runtime version 3.0.0.

Core runtime

  • stdio MCP router and multi-server aggregator
  • native tools:
    • docs_architect
    • skill_creator
    • resolve_workflow
    • search_skills
    • get_skill_help
    • discover_tools
    • proxy_server_status
    • refresh_skills
    • fetch_skill
    • execute_skill
  • MCP resources: evokore://server/status, evokore://server/config, evokore://skills/categories
  • MCP prompts: resolve-workflow, skill-help, server-overview
  • proxied child-server model through mcp.config.json
  • namespaced proxied tools using ${serverId}_${tool.name}

Major operator capabilities

  • dynamic tool discovery with session-scoped activation
  • recursive skill indexing with metadata-aware search
  • semantic workflow resolution with reranking and Why matched: output
  • HITL approval enforcement through _evokore_approval_token
  • canonical hook entrypoints and shared fail-safe bootstrap
  • standalone VoiceSidecar with persona-aware hook forwarding
  • MCP SDK tool annotations (readOnlyHint, destructiveHint, etc.)
  • HTTP client transport support for child servers
  • skill hot-reload via refresh_skills and optional filesystem watcher
  • configurable per-server and per-tool rate limiting (token bucket)
  • session dashboard for replay and evidence timeline viewing
  • interactive HITL approval web UI
  • optional pre-session repo audit hook
  • skill versioning with dependency validation
  • remote skill registry and fetch_skill tool
  • skill execution sandbox with execute_skill tool
  • Supabase MCP server as a proxied child with tiered permissions
  • RBAC permissions model with admin, developer, and readonly roles

Continuity and repo-operations layer

  • canonical session manifest under ~/.evokore/sessions/{sessionId}.json
  • managed Claude memory sync via npm run memory:sync
  • manifest-backed status line through scripts/status.js
  • repo-state preflight via:
npm run repo:audit
npm run repo:audit -- --json

Testing

  • 72 test files, 179 tests via vitest
  • parallel execution with watch mode support
  • structured output with vitest run

Validation baseline

Primary commands:

npm run build
npm test
npm run repo:audit

Targeted operator checks:

  • node test-session-continuity-validation.js
  • node test-auto-memory-validation.js
  • node test-status-line-validation.js
  • node test-damage-control-validation.js
  • node test-repo-state-audit-validation.js
  • node test-version-contract-consistency.js

Recommended doc path

  1. ../README.md
  2. SETUP.md
  3. USAGE.md
  4. ARCHITECTURE.md
  5. TESTING_AND_VALIDATION.md
  6. CHANGELOG.md