Skip to content

Changelog

All notable changes to the Knowledge Plugin will be documented in this file.

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

Unreleased

0.1.0-beta - 2026-03-03

Added

  • First beta release milestone — infrastructure and features stable for external testing
  • All features from alpha cycle (v0.0.9 through v0.0.11-alpha) integrated and verified
  • Comprehensive pre-beta test suite (113 tests, 9 suites) validating all systems

Changed

  • Version number bumped from v0.0.11-alpha to v0.1.0-beta to reflect stability
  • Updated command count (25), skills count (6), subagents count (3) across documentation

Fixed

  • Pre-commit sanitization hook now bash 3.2 compatible (macOS default) — no longer fails silently
  • marketplace.json version two releases behind (0.0.9-alpha → 0.1.0-beta)
  • mcp-server/src/index.ts hardcoded version "1.0.0" never updated from scaffold
  • marketplace.json typo: knowedge-graphknowledge-graph
  • README.md typo: avaliableavailable

Documentation

  • Added v0.1.0-beta section to ROADMAP.md
  • Synced CHANGELOG, ROADMAP, README to v0.1.0-beta

Important Note

This is a beta release. API and behavior may change before a stable release.


[0.0.10.4-alpha] - 2026-03-01

Fixed

  • MCP Server Auto-Install on Plugin Cache Miss
    • hooks-master.sh Section 1 previously skipped npm install when dist/index.js existed, even if node_modules/ was absent (common after marketplace plugin install)
    • Split check into NEEDS_INSTALL (node_modules/@modelcontextprotocol missing) and NEEDS_BUILD (dist/index.js missing); each triggers independently
    • Plugin installs with pre-built dist/ now auto-install deps on first session start
    • When both exist: Section 1 is a no-op (no performance regression)

Branch: v0.0.10.4-fix-mcp-missing-node-modules


[0.0.10-alpha] - 2026-02-27

Added

  • Skills System (v0.0.10.1)

    • 5 auto-triggered context providers: lesson-capture (bug solved), kg-recall (past decisions), session-wrap (context limits), adr-guide (architecture decisions), gov-execute-plan (plan execution)
    • Skill auto-surfaces suggestions when trigger conditions detected in conversation
    • Reduces context overhead by ~46.9% via lazy-loading
  • Subagents for Heavy-Lift Tasks (v0.0.10.1)

    • knowledge-extractor — Read-only parsing of large files; presents findings for user approval before writing
    • session-documenter — Git archaeology for session summaries with conventional commit format
    • Both operate in approval-gated mode; never auto-writes without user confirmation
  • Optional KG Backfill During Init (v0.0.10.2)

    • New Step 1.10 in /kmgraph:init: "Would you like to backfill the knowledge graph from existing project context? [y/N]"
    • If yes: invokes knowledge-extractor to scan README, CHANGELOG, lessons-learned/, decisions/, and chat-history/
    • Extracts candidates and presents for user review before creating entries
    • Enables new users to inherit institutional knowledge immediately
  • Handoff Command (v0.0.10.3)

    • New /kmgraph:handoff — Generate comprehensive project transition documentation
    • Creates 5 documents: START-HERE (current state), DOCUMENTATION-MAP (file inventory), SESSION-COMPILATION (recent work), OPEN-ISSUES (blockers), ARCHITECTURE-SNAPSHOT (codebase structure)
    • Purpose: Enable seamless knowledge transfer before context limits or developer transitions
  • Delegation Options for Heavy Reads (v0.0.10.3)

    • Updated /kmgraph:extract-chat, /kmgraph:session-summary, /kmgraph:update-graph with guidance
    • When processing 10+ sessions or 50+ KB, suggests delegating to appropriate subagent
    • Reduces peak context usage for large operations
  • Documentation Configuration (v0.0.10.3)

    • Moved CHANGELOG.md to top-level mkdocs navigation (was nested under Contributing)
    • Added LinkedIn icon to header (between GitHub and search) via custom theme override
    • Updated COMMAND-GUIDE.md with handoff section and new delegation patterns
    • Updated GETTING-STARTED.md with skills and subagents explanation and trigger table

Changed

  • Commands Reference

    • Added /kmgraph:handoff (new command)
    • Command count updated: 22 → 23 total commands
    • Install instructions now reference kmgraph namespace consistently

Documentation

  • Added "Skills and Subagents" section to GETTING-STARTED.md with trigger tables
  • Added "Project Transitions & Onboarding" section to COMMAND-GUIDE.md
  • Enhanced CHEAT-SHEET.md with delegation syntax examples
  • Updated CLAUDE.md with Skills, Subagents, and Commit Format sections

Commits across v0.0.10.0 through v0.0.10.3:

  • v0.0.10.0: Cleanup and workflow consolidation
  • v0.0.10.1: Skills (5), Subagents (2), and KG backfill scaffolding
  • v0.0.10.2: Integrated init backfill option (knowledge-extractor powered)
  • v0.0.10.3: Handoff command, delegation blocks, documentation updates, theme customization

[0.0.9-alpha] - 2026-02-27

Added

  • Plugin Infrastructure & Onboarding
    • Added CLAUDE.md to define project architecture, versioning rules, and strict AI constraints
    • Re-introduced mcpToolSearch: true in settings to enable lazy-loading and reduce token overhead
    • Step 0 ("Permissions") and Step 0.5 ("Migration Check") added to INSTALL.md

Changed

  • Complete Namespace Migration
    • Renamed namespace across all code, manifest, and documentation from /kg-sis: to /kmgraph:
    • kg-sis plugin disabled in settings; enabled kmgraph as the only active extension identifier
      • Uninstaller included
      • Might require clearing of local cache
  • Consolidated Automation Hooks
    • Replaced 3 separate shell scripts with single hooks-master.sh invoking 3 isolated sections (config, lessons, memory)
    • Updated hooks.json to reduce plugin load overhead

Fixed

  • Hook Security Audit (ADR-012)
    • Applied word-splitting protections (quoted subshells) in memory-diff-check.sh
    • Validated strict avoidance of eval, network requests, and code-altering operations in all hooks
    • Ensured MEMORY.md limits conform exactly to ADR-004 logic (2,000 token limit)

Documentation

  • Created ADR-012: Hook Security Model defining rules for plugin script execution
  • Added non-Claude platform instructions to GETTING-STARTED.md and COMMAND-GUIDE.md
  • Fixed lingering references to legacy namespace in NAVIGATION-INDEX.md and UI test files

[0.0.8.7-alpha] - 2026-02-22

Added

  • Manual Documentation Updates & Security Fixes
    • ROADMAP version history table: Complete chronological record v0.0.1-alpha through v0.0.8.6-alpha
    • ROADMAP footer: Updated to reflect v0.0.8.6-alpha current release

Fixed

  • Security: npm audit vulnerabilities → 0 vulns
    • Fixed ajv ReDoS vulnerability (GHSA-2g4f-4pwh-qvx6) in MCP server dependencies
    • Fixed hono timing comparison hardening (GHSA-gq3j-xvxp-8hrf) in MCP server dependencies
    • Rebuilt mcp-server/dist/ with patched packages
    • Updated mcp-server/package-lock.json with fixed versions

Documentation

  • Removed hardcoded version numbers to prevent docs becoming stale:

    • Changed "22 slash commands" → "slash commands" in index.md
    • Changed "7 MCP tools" → "MCP tools" in index.md
  • Added clarity to Four Pillars reference: "Learn about the Four Pillars this project was built on"

  • Simplified GETTING-STARTED.md heading format (removed "Path A" prefix)
  • Updated Getting Started card: Specified "local IDE CLI coding assistant" for clarity on platform scope

Branch: v0.0.8.7-alpha-manual-updates Commits:

  • ca59e184 - Docs: Remove hardcoded version numbers + clarifications
  • 9830f8aa - Build: Fix npm security vulnerabilities
  • 949b04c1 - Docs: ROADMAP version history table
  • 0f8c19b7 - Docs: ROADMAP MkDocs customization section
  • 4e91d8e7 - Docs: CHANGELOG backfill v0.0.7-alpha through v0.0.8.4-alpha
  • c92e9f7e - Docs: Getting Started card "local IDE CLI coding assistant"

[0.0.8.6-alpha] - 2026-02-22

Added

  • MkDocs Material Theme Customization (Phases 1-3)

    • Material theme v9.7.0+ with 10+ navigation features enabled
    • Dark mode (slate scheme) as default with light mode fallback
    • Sticky navigation tabs (navigation.tabs.sticky)
    • Breadcrumbs above page titles (navigation.path)
    • Footer navigation with Next/Previous buttons (navigation.footer)
    • Integrated Table of Contents in left sidebar (toc.integrate + toc.follow)
    • Copy buttons on all code blocks (content.code.copy)
    • Search plugin configuration with autocomplete, highlighting, and sharing
  • Custom CSS Styling (400+ lines)

    • Typography: Inter and JetBrains Mono from Google Fonts
    • Dark mode colors: Navy primary (#1a1a2e), cyan accent (#00d2ff)
    • Light mode colors: Blue primary (#003d82), orange accent (#ff6b35)
    • Glassmorphism header effect with backdrop blur (dark mode only)
    • WCAG AA contrast compliance for all color combinations
    • Enhanced code blocks, tables, admonitions, and search box styling
    • Print media support (hides navigation for exports)
  • Page Restructuring & Visual Enhancements

    • Grid cards on index.md and GETTING-STARTED.md for visual navigation
    • Tabbed interface in COMMAND-GUIDE.md (6 command categories)
    • Mermaid diagrams: "Knowledge Capture Pipeline" (GETTING-STARTED.md)
    • Mermaid diagrams: "Four Pillars Relationships" (CONCEPTS.md)
    • All diagrams include accessibility attributes (accTitle, accDescr)
    • Neutral mermaid theme for proper rendering in both color schemes
  • Additional Experience Plugins

    • mkdocs-git-revision-date-localized-plugin — "Last updated" timestamps on all pages
    • mkdocs-glightbox — Lightbox image/diagram viewing
    • mkdocs-minify-plugin — Asset compression for snappy performance
    • mkdocs-roamlinks-plugin — WikiLink support for knowledge entries
  • Social Links & Copyright

    • GitHub: https://github.com/technomensch
    • LinkedIn: https://www.linkedin.com/in/marckaplan/
    • Copyright: "Staying in Sync"

Fixed

  • Mermaid diagram rendering in dark mode (removed hardcoded colors, adopted neutral theme)
  • Light mode header styling (restored Material theme defaults)
  • Grid card links accessibility (ensured descriptive text, no "click here")

Documentation

  • Updated STYLE-GUIDE.md with blockquote vs. admonition format guidance
  • Moved FAQ under Commands section in mkdocs.yml navigation
  • Added comprehensive Section 508 compliance documentation in STYLE-GUIDE.md

Technical

  • Updated requirements.txt with mkdocs-material>=9.7.0 and 4 plugins
  • mkdocs.yml: 15+ navigation features, plugin configuration, theme palette setup
  • Custom stylesheet: docs/stylesheets/extra.css (400+ lines)
  • No core document rewrites or file splitting (MkDocs rendering enhancements only)

Version: 0.0.8.4-alpha → 0.0.8.6-alpha

[0.0.6-alpha] - 2026-02-17

Added

  • Root package.json with files allowlist — implements npm-standard distribution hygiene so marketplace-installed plugin excludes developer-only content:
    • docs/ (plugin developer's knowledge graph: decisions, lessons, KG entries)
    • tests/ (internal test suite)
    • Root development files (ROADMAP.md, etc.) docs/ directory remains in git unchanged; no path changes to commands or scripts.

Fixed

  • Stale kg-config.json path: knowledge-graph-plugin/docsknowledge-graph/docs (repo was renamed in v0.0.3 but local config was never updated)
  • Stale GitHub URLs: updated knowledge-graph-pluginknowledge-graph in CHANGELOG footer, ROADMAP feedback links, README install example, tests/README, and scripts

Documentation

  • Added developer vs. distribution table to README.md
  • Updated ROADMAP.md with v0.0.6-alpha section

Version: 0.0.5-alpha → 0.0.6-alpha

[0.0.5-alpha] - 2026-02-17

Added

  • /kmgraph:start-issue-tracking command (19th command) — Full issue initialization workflow, fully ported from prior project and sanitized for cross-project portability and LLM-platform-agnostic use. Features:
    • Auto-detects parent branch, version from git tag, issue type, and next issue number from existing issues/ directory
    • Smart defaults reduce interactive prompts to 1 (issue description only)
    • Creates structured directory under {active_kg_path}/issues/{number}-{slug}/
    • Generates issue.md with full metadata (title, type, branch, version, date, scope)
    • Git branch creation: git checkout -b issue/{number}-{slug}
    • Knowledge graph synchronization via /kmgraph:update-issue-plan
    • Integrates with /kmgraph:link-issue and /kmgraph:meta-issue
    • No project-specific dependencies; uses KG config for all path resolution

Fixed

  • .gitignore inline comments on pattern lines (3 paths were silently not being ignored because git does not support inline comments on pattern lines)
  • Truncated marketplace slug stayinginsync-knowledge-grap (missing trailing h) in .claude/settings.json and plugin cache settings.json — caused plugin-not-found errors on every session start
  • Dangling /kmgraph:start-issue-tracking references in commands/update-issue-plan.md (lines ~61 and ~203) now resolve to the newly created command
  • First SessionStart hook entry (check-memory.sh) missing comment field
  • Session-summary command markdown template embedded as raw prose instead of fenced code block, causing visual ambiguity between instruction and template content
  • Standardized command frontmatter: removed name field from recall.md, list.md, and session-summary.md for consistency with all other 16 commands

Removed

  • Empty mcp-server/.claude-plugin/ artifact directory (leftover from refactoring, risked being parsed as a nested plugin by plugin discovery tools)
  • Orphaned root-level node_modules/ directory (no root package.json exists; packages were installed by mistake at an earlier point)

Documentation

  • Added docs/lessons-learned/architecture/.gitkeep and docs/lessons-learned/patterns/.gitkeep to preserve empty tracked directories
  • Updated ROADMAP.md with v0.0.5-alpha section
  • Updated README.md version, status line, and command count (18 → 19)
  • Added implementation plan: docs/plans/v0.0.5-alpha-plan.md

Version: 0.0.4-alpha → 0.0.5-alpha

[0.0.4-alpha] - 2026-02-16

Added

  • /kmgraph:restore-memory Command - Restore archived MEMORY.md entries
    • Fuzzy search by entry title using fuzzy-search-archive.sh helper script
    • Restore by entry ID/index with --id flag
    • List all archived entries with --list flag
    • Preview entry content before restoring
    • Target section selection with --section flag (auto-detect or user-specified)
    • Dry-run mode with --dry-run flag for previewing without writing
    • Token limit checking (blocks if would exceed 2,000 tokens, warns if > 1,500)
    • Archive log restoration tracking (marks entries as "Restored: YYYY-MM-DD")
    • Commits both MEMORY.md and MEMORY-archive.md with descriptive message
  • Fuzzy Search Script - scripts/fuzzy-search-archive.sh
    • Four-tier ranking strategy: exact match, starts-with, contains-all words, contains-any word
    • Case-insensitive search with word-based fuzzy matching
    • Returns ranked list of matching entry IDs and titles
  • Architecture Decision Record - docs/decisions/ADR-005-defer-memory-rules-engine.md
    • Documents decision to defer rules engine and smart summarization to v0.0.5-alpha
    • Analyzes three options: rules+restore (medium scope), full automation (all features), restore only (minimal scope)
    • Rationale: Archive without restore is incomplete UX, rules need real-world patterns, maintain velocity

Changed

  • Version: 0.0.3-alpha → 0.0.4-alpha
  • Command Count: 17 → 18 (added restore-memory)
  • /kmgraph:archive-memory Command - Enhanced with restoration tracking
    • Archive log now shows restoration timestamps: "[Restored: YYYY-MM-DD]"
    • Restored entries remain in archive for historical record
    • Documents restore workflow and manual restoration process
  • knowledge-graph-usage skill - Added restore workflow documentation
    • When to restore archived entries (context needed for current work)
    • Restore vs archive decision criteria
    • Integration with archive-memory command

Documentation

  • Added implementation plan: docs/plans/v0.0.4-alpha-plan.md
    • Complete 3-phase implementation breakdown
    • 27 verification checkboxes across 4 categories
    • Timeline estimation (2-3 days)
  • Updated ROADMAP.md with v0.0.4-alpha section
  • Updated README.md command count and status

Deferred

  • MEMORY.md auto-sync rules engine (deferred to v0.0.5-alpha)
    • YAML-based pattern matching for automated sync decisions
    • Global defaults + per-KG overrides
    • Confidence scoring system
  • Smart summarization (deferred to v0.0.5-alpha)
    • LLM-powered entry consolidation
    • Batch processing or on-demand
    • Merge similar entries strategy

[0.0.3-alpha] - 2026-02-16

Added

  • /kmgraph:archive-memory Command - Archive stale MEMORY.md entries to prevent bloat
    • Token-based staleness detection (90-day threshold, customizable)
    • Moves stale entries to MEMORY-archive.md with archive log
    • Shows tokens freed and current size after archival
    • Dry-run mode for previewing without writing
  • Autonomous Triggering in Knowledge-Graph-Usage Skill
    • After lesson capture: Suggests /kmgraph:update-graph immediately
    • After significant commits: Detects fix/debug/pattern keywords, suggests capture within 30 minutes
    • Before problem-solving: Suggests /kmgraph:recall to check existing knowledge
  • Post-Commit Hook Template - Detects lesson-worthy commits
    • Located in core/examples-hooks/post-commit-lesson-suggestion
    • Triggers on keywords: fix, solved, debug, implement, refactor, pattern, architecture
    • Optional installation via /kmgraph:init wizard (default: no)
  • SessionStart Hooks - Three hooks for enhanced context
    • recent-lessons.sh - Displays lessons modified in last 7 days
    • memory-diff-check.sh - Notifies of MEMORY.md changes since last session
    • Both scoped to active KG, silent when no changes
  • Duplicate Detection Pre-Flight - Step 1.1 in capture-lesson
    • Searches for similar lessons before content gathering
    • Offers merge (update existing), link (create with reference), or proceed (new)
    • Prevents knowledge fragmentation

Changed

  • Version: 0.0.2-alpha → 0.0.3-alpha
  • Command Count: 16 → 17 (added archive-memory)
  • MEMORY.md Limits: Line-based (250/300) → Token-based (1,500/2,000)
    • Token estimation: word_count × 1.3
    • Soft limit: 1,500 tokens (warning, sync continues)
    • Hard limit: 2,000 tokens (blocks MEMORY.md updates, suggests archive)
    • Replaced all line-based references in update-graph.md Step 7 and sync-all.md
  • capture-lesson.md Step 4.6 - Structured choice UI
    • "Extract now (recommended)" - Inline update-graph execution
    • "Manual later" - Deferred extraction
    • "Skip" - Batch via sync-all
  • update-graph.md - Enhanced --auto flag behavior
    • Returns structured quality feedback when called from capture-lesson
    • Added --edit-entry flag for user review before saving
  • knowledge-graph-usage skill - Added duplicate detection guidance (~150 words)
    • Search strategy before capturing
    • Merge vs create new decision criteria

Fixed

  • Token-based size limits more accurate than line-based (short vs long lines)
  • MEMORY.md bloat prevention via archival system
  • Knowledge fragmentation via duplicate detection

Documentation

  • Plan: docs/plans/v0.0.3-alpha-plan.md (257 lines, consolidated from 1,174)
  • ROADMAP: v0.0.3-alpha section added with 3-phase breakdown
  • Verification: All Phase 1, 2, 3 checkboxes marked complete

Deferred to v0.0.4-alpha

  • MEMORY.md auto-sync rules engine (YAML rules, confidence scoring)
  • Smart summarization (LLM-powered entry consolidation)
  • /kmgraph:restore-memory command (restore archived entries by ID)
  • Per-KG config directories with memory-sync-rules.yaml

[0.0.2-alpha] - 2026-02-16

Added

  • Knowledge Graph Usage Skill - Autonomous guidance for knowledge capture
    • 1,900-word lean SKILL.md with progressive disclosure
    • 5,800-word capture-patterns.md reference (problem-solution, architectural, meta-issue patterns)
    • 6,200-word command-workflows.md reference (10 detailed workflow patterns)
    • Triggers on phrases: "documenting lessons", "institutional memory", "we solved this before"
    • Proactive recognition of recurring problems and valuable insights
  • Plugin Knowledge Graph - Plugin now documents itself (dogfooding)
    • Initialized KG in docs/ with categories: architecture, debugging, patterns
    • Selective git strategy (commit shareable, gitignore personal notes)
    • First lesson captured: namespace-visibility-shadow-command-failure.md
    • Master index with chronological and tag-based navigation
  • Marketplace Branding - Changed identifier from "(knowledge)" to "(tm-sis)"
    • Represents "technomensch-stayinginsync" publisher identity
    • Updated marketplace.json with new branding
    • README documentation of marketplace strategy

Changed

  • Version: 0.0.1-alpha → 0.0.2-alpha
  • Command Filenames: Removed knowledge- prefix from all 16 command files
    • knowledge-status.mdstatus.md (all commands renamed)
    • Marketplace installation shows namespace correctly regardless of filename
    • Cleaner, more maintainable filenames
    • Git history preserved via rename detection
  • README: Corrected command count from 17 to 16 (accurate count)
  • README: Updated namespace documentation to reflect marketplace behavior
    • Documents two-location sync requirement for local testing
    • Explains Distribution Mode namespace handling
    • References captured lessons for detailed workflow
  • .gitignore: Added selective KG strategy rules
    • Gitignore: docs/plans/, docs/sessions/, docs/chat-history/, docs/lessons-learned/debugging/
    • Commit: docs/lessons-learned/architecture/, docs/lessons-learned/patterns/, docs/lessons-learned/process/

Fixed

  • Filename typo: knowledge-updat-issue-plan.mdknowledge-update-issue-plan.md

Documentation

  • Lesson 1: Shadow command strategy failed with Gemini (cross-LLM incompatibility)
    • File: docs/lessons-learned/debugging/namespace-visibility-shadow-command-failure.md
    • Documented file prefix workaround as cross-LLM compatible solution
    • Updated with marketplace discovery (namespace works correctly regardless of prefix)
    • Cross-references local marketplace testing workflow lesson
  • Lesson 2: Local marketplace testing requires two-location sync
    • File: docs/lessons-learned/process/local-marketplace-testing-workflow.md
    • Documents development directory vs marketplace cache locations
    • Provides rsync automation script for sync workflow
    • Explains Distribution Mode namespace behavior
  • Master Index: Updated with 2 lessons total (debugging + process categories)
    • Chronological index with date-based navigation
    • Tag index with 9 unique tags (#testing, #marketplace, #plugin-development, etc.)
  • Updated plugin validation criteria checklist in v0.0.2-validate-plugin.md plan

Validation

  • Plugin-validator: PASS with 0 critical issues
  • All 16 commands validated
  • Skill validated with proper progressive disclosure
  • MCP server validated (7 tools, 2 resources)

0.0.1-alpha - 2026-02-16

Added

  • Initial alpha release of Knowledge Plugin for Claude Code
  • 16 commands for knowledge capture, recall, sync, and management
  • Multi-KG support with per-category git strategies
  • Git metadata tracking in lesson/ADR YAML frontmatter
  • MCP server (7 tools + 2 resources) for cross-platform use
  • Platform-agnostic core for non-Claude users (Cursor, Continue.dev, Aider, local LLMs)
  • ~30 generalized examples + 10 documentation files
  • SessionStart hook for MEMORY.md staleness detection
  • Python chat extraction scripts (Claude + Gemini)
  • Meta-issue tracking system for complex multi-attempt problems
  • Bidirectional KG ↔ MEMORY.md synchronization
  • Privacy-focused sanitization tools and documentation

Changed

  • ARCHITECTURAL DECISION: Migrated from skills/ to commands/ directory
    • Commands provide manual invocation (not autonomous)
    • Full workflow loading (not lazy-loaded)
    • Better suited for deterministic knowledge operations
  • Updated plugin metadata (version 0.0.1-alpha, email, repository, license, keywords)
  • Updated README with commands vs skills architecture documentation
  • All command name: fields now include knowledge: namespace prefix for autocomplete

Commands (renamed to "kmgraph" in v.0.0.10-alpha)

  • /kmgraph:init - Initialize new knowledge graph with wizard
  • /kmgraph:list - Display all configured knowledge graphs
  • /kmgraph:switch - Change active knowledge graph
  • /kmgraph:add-category - Add category to existing KG
  • /kmgraph:configure-sanitization - Set up pre-commit hooks for sensitive data
  • /kmgraph:check-sensitive - Scan KG for potentially sensitive information
  • /kmgraph:link-issue - Link lesson to GitHub issue with bidirectional references
  • /kmgraph:status - Display active KG status and quick reference
  • /kmgraph:capture-lesson - Document lessons with git metadata
  • /kmgraph:recall - Search across all KG systems
  • /kmgraph:update-graph - Extract insights from lessons to KG
  • /kmgraph:sync-all - Automated knowledge sync pipeline
  • /kmgraph:update-issue-plan - Sync KG → plan → issue → GitHub
  • /kmgraph:session-summary - Auto-document work sessions
  • /kmgraph:extract-chat - Extract chat history from Claude/Gemini logs
  • /kmgraph:meta-issue - Initialize meta-issue tracking for complex problems

MCP Tools

  • kg_config_init - Create KG directory structure + config entry
  • kg_config_list - Read and return all KGs from config
  • kg_config_switch - Update active KG in config
  • kg_config_add_category - Add category directory + update config
  • kg_search - Full-text search across KG files
  • kg_scaffold - Create file from template with variable substitution
  • kg_check_sensitive - Scan files against regex patterns

MCP Resources

  • kg://config - Current kg-config.json contents (read-only)
  • kg://templates/{name} - Template files from core/templates/

[Features under consideration for Future Release]

Added

  • Initial release of Knowledge Plugin for Claude Code
  • 16 skills for knowledge capture, recall, sync, and management
  • Multi-KG support with per-category git strategies
  • Git metadata tracking in lesson/ADR YAML frontmatter
  • MCP server (7 tools + 2 resources) for cross-platform use
  • Platform-agnostic core for non-Claude users (Cursor, Continue.dev, Aider, local LLMs)
  • ~30 generalized examples + 10 documentation files
  • SessionStart hook for MEMORY.md staleness detection
  • Python chat extraction scripts (Claude + Gemini)
  • Meta-issue tracking system for complex multi-attempt problems
  • Bidirectional KG ↔ MEMORY.md synchronization
  • Privacy-focused sanitization tools and documentation

Planned Features (v1.0.0)

  • TBD