Skip to content

v8.0.0 - Memory Journal MCP v8.0.0

Choose a tag to compare

View all tags
@neverinfamous neverinfamous released this 31 May 22:02
· 5 commits to main since this release
v8.0.0
fcab2d7
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Memory Journal MCP v8.0.0

Highlights

  • Team Flag Management (Hush Protocol): Complete parity with team_list_flags, team_update_flag, and team_get_flag_analytics.
  • Code Mode Enhancements: Major improvements to the sandbox with runtime introspection, alias enhancements, and strict protocol meta-object boundaries to prevent abuse.
  • Agent Capabilities: 20+ new skills added including adversarial planning, and expanded documentation and tracking metrics.
  • Performance & Cleanup: Removed redundant endpoints and optimized dynamically generated TypeScript declarations saving tokens.

Added

  • admin: project_number, significance_type, and GitHub metadata fields to update_entry and team_update_entry
  • analytics: On-read computation with 60s TTL cache, startup snapshot seeding, and live fallback for memory://insights/digest
  • auto-prune: Importance-based garbage collection for old, low-importance entries via CLI flags and environment variables
  • briefing: Server version, resource/tool counts, test health indicators, workspace paths, git working tree status, code-map availability, 📊 memory://metrics/summary breadcrumb, and unreleased change summary to memory://briefing output
  • codemode: Enhanced sandbox with context/readonly parameters, runtime schema introspection, progress notification bindings, and additional aliases
  • docs: Missing CLI flags, environment variables, auto-prune documentation, and "Customizing the Session Briefing" sections
  • errors: Valid enum value lists to Zod validation error messages
  • github: Dependabot configuration scoped to github-actions only for passive version update notifications
  • instructions: 14 new EntryType values, architecture/security significance types, missing tool parameters, and HELP_CONTENT export
  • metrics: Dynamic tracking for deprecation warnings (MetricsAccumulator.recordDeprecationWarning)
  • prompts: adversarial-plan-review and flag-dashboard prompts for multi-pass planning and flag triage
  • resources: memory://flags/history resource showing recently resolved flags with resolution details and average time-to-resolution metrics
  • scripts: Added test:scheduler and test-progress.mjs for E2E and integration testing
  • skills: Added 20+ skills including official vendor integrations, adversarial auditing, and docs-marketer
  • team: Added team_list_flags, team_update_flag, and team_get_flag_analytics tools for flag management
  • tests: verify-schemas.mjs script to programmatically validate outputSchema definitions
  • ci/cd: Added Docker smoke test job to docker-publish.yml that verifies binary loading and HTTP transport starts successfully before images are pushed to the registry
  • ci/cd: Added Docker smoke test job to lint-and-test.yml for PR-level validation, catching dependency resolution failures before merge
  • tests: Added build-externals.test.ts invariant test ensuring every tsup.config.ts external entry has a corresponding production dependency

Changed

  • briefing: Enhanced rendering with filtered milestones, capability statuses, and prioritized CI workflows
  • codemode: Optimized dynamically generated TypeScript declarations and removed redundant strings to save tokens
  • deps: Updated npm dependencies including commander to ^15.0.0 and eslint to 10.4.1
  • docker: Bumped Node.js base image to 26.2.0-alpine (#563)
  • docs: Highlighted auto-prune, standardized README.md layout, and simplified agent briefing instructions
  • instructions: Refactored server-instructions.md into modular directory, saving ~700 tokens, and updated help handlers
  • prompts: Enhanced flag-dashboard and surfaced team flags in standup/retro prompts
  • resources: Changed memory://briefing and memory://briefing/{repo} to return text/markdown directly instead of JSON objects for token efficiency
  • schemas: Enforced strict project_number validation across Core and Team tools with formatted validation errors
  • skills: Restructured large skills, expanded mcp-builder, and upgraded adversarial-planner to embed gh copilot scans
  • tests: Standardized 42+ prompt files, added artifact cleanup, and clarified Code Mode restrictions

Removed

  • instructions: gotchas.md, codemode.md, and server-access.md static help files to reduce agent distraction and token usage
  • resources: Redundant memory://briefing-message and memory://briefing-message/{repo} endpoints

Fixed

  • annotations: Added missing destructiveHint: false to 62 tools and idempotentHint: false to delete_entry
  • briefing: Fixed dynamic context routing, deduplicated entry previews, and rendering cutoff issues
  • codemode: Comprehensive hallucination interception for API aliases, missing parameters, and unsupported cross-server database queries
  • codemode: Expanded global sandbox bindings, proxy mappings, and parameter coercions for robust agent interactions
  • codemode: Improved error handling and validation for legacy payloads, syntax errors, and missing fields
  • core: Enforced strict validation for significance types and fixed auto_context persistence mapping
  • docs: Synchronized code-map.md, README.md, environment variables, and tool counts across documentation
  • github: Fixed parameter hallucinations in repo insights, payload truncation flags, and repository resolution fallbacks
  • instructions: Fixed hallucinated JSON parsing instructions, corrected tool schemas, and restored missing gotchas.md help content
  • io: Fixed export_entries ignoring limits and increased ceiling to 5000
  • progress: Fixed step numbering gaps and missing notifications in export and backup sequences
  • relationships: Fixed visualize_relationships returning null when empty, and intercepted hallucinated schema parameters
  • resources: Fixed memory://help failing to extract parameters from ZodPipelines
  • repo: Mapped repo_name fallback to repo within mj_execute_code sanitization to prevent context loss
  • scripts: Improved E2E scheduler test scripts and help pointer mapping logic
  • server: Fixed structured error responses for Zod validations and payload limits, preventing raw JSON-RPC leakages
  • skills: Remediated frontmatter, database triggers, and stub implementations across migrated skills
  • team: Fixed parameter parsing, path traversal checks, schema errors, and project_number filtering across team tools
  • tests: Fixed environment resolution, hallucinated expectations, and missing parameters across 26 test scripts

Security

  • codemode: Nullified Proxy/Reflect/Symbol constructors and added frozen built-in prototypes to VM sandbox to prevent meta-object protocol abuse
  • docs: Documented engine-level Code Mode sandbox restrictions and RPC allowlists in SECURITY.md and README.md
  • skills: Bumped qs dependency in gitlab skill to resolve a remotely triggerable DoS vulnerability

Full Changelog: v7.7.1...v8.0.0

# Update via npm
npm install -g memory-journal-mcp@latest

# Update via Docker
docker pull writenotenow/memory-journal-mcp:latest
Assets 2
Loading