dino

Automatic context engineering for Claude Code

       __
      (_ \
  .-^^^  /
 /       /
<__|-|__|

dino brings structured, context-aware development to Claude Code. It automatically maintains session state, provides 75 MCP tools, and enables intelligent context retrieval with real-time statusline display.

v17.6.0: Session Context Integration - Skills and MCP tools now automatically load session context (focus, phase, blockers) - See CHANGELOG.md for details

Features

v17.6.0 - Session Context Integration

• Automatic Session Context Loading - Skills and MCP tools now context-aware by default

  • MCP Tools Enhanced - 10 tools auto-load focusArea, phase, blockers from .dino/session.md
  • Session Context Wrapper - withSessionContext() wrapper provides consistent session access
  • Forked Skills Fixed - 5 skills with Step 0 context recovery prevent losing context in fork mode
  • Session-Aware Skills - 7 skills explicitly read session.md before executing work
  • Context-Aware Functions - 5 TypeScript functions accept optional session context parameters

• Benefits

  • No Repeated Questions - Skills reuse context already in session instead of asking user
  • Focus-Aware Operations - MCP tools automatically scope to current focus area
  • Better Handoffs - Forked skills recover context from session, ralph-gate, or handoff files
  • Smarter Defaults - Functions fall back to session.md when context not explicitly provided

v17.4.0 - Research Validation System

• Research Validation - Automatic validation of package recommendations before installation

  • Pre-Install Hook - Intercepts package manager commands (bun add, npm install, yarn add)
  • Security Scanning - npm audit integration for vulnerability detection
  • Quality Checks - Package size, download stats, maintenance status
  • Blocking Rules - Critical/high vulnerabilities block installation
  • Rate Limiting - Exponential backoff to respect API limits
  • Caching - 1-hour TTL for metadata, 24-hour for size info

• New Skill: /dino.validate-deps - Manual dependency validation

  • Validates all dependencies in package.json
  • --security-only flag for quick security checks
  • Structured reports with actionable recommendations

• New MCP Tools - 3 validation tools

  • dino_validate_package - Full package validation
  • dino_validate_deps - Batch validation
  • dino_security_scan - Security-only scan

• Ralph Gate Integration - Validation in three-agent gating system

  • Scout identifies packages in PRD
  • Auditor blocks on security issues
  • Results saved to validation report

v17.2.0 - TypeScript Status Tracking

• TypeScript Status Tracking - Automatic TypeScript error detection in session state
  • Post-Edit Hook - Runs tsc --noEmit after Edit/Write operations
  • Session State Updates - TypeScript status tracked in session.md
  • Error Reporting - File paths, line numbers, and error messages captured
  • Hook Integration - Wired to PostToolUse hook via post-typescript script
  • Incremental Compilation - Uses --incremental flag for faster checks
  • Platform Support - Windows (powershell), macOS/Linux (bash)

v17.1.0 - Settings Validator & Config Auto-Repair

• Settings Validation - Comprehensive validation for Claude Code configuration

  • Schema Validation - Validates .claude/settings.json structure
  • Hook Validation - Checks hook event types, matchers, command paths
  • StatusLine Validation - Validates statusLine configuration
  • CLI Command - dino init --validate for manual validation
  • Session-Start Check - Auto-validates config on session start
  • Error Reporting - Detailed validation errors with suggestions

• Config Auto-Repair - Automatic fixing of common configuration issues

  • Repair Engine - Fixes invalid settings automatically
  • Hook Repairs - Corrects missing "hooks" wrappers, invalid matchers
  • Path Repairs - Expands $CLAUDE_PROJECT_DIR to absolute paths
  • Type Repairs - Fixes incorrect "type" field values
  • Backup System - Creates .claude/settings.json.backup before repairs
  • Validation Integration - Runs after validation detects issues

• Benefits

  • Prevent Hook Failures - Catches configuration errors before they break hooks
  • Auto-Fix Common Issues - No manual editing required for known problems
  • Session Reliability - Ensures hooks execute correctly on session start
  • Better Error Messages - Clear explanations of what's wrong and how to fix it

v16.2.0 - Smart Init Mode & Safe Updates

• Smart Update Mode - Safe, non-destructive updates that preserve customizations

  • session.md Preservation - Never regenerates session.md in update mode (preserves context)
  • Selective CLAUDE.md Updates - Only updates dino sections, preserves user content
  • Customization Detection - Automatically skips customized files, updates only unchanged files
  • Deprecated File Cleanup - Centralized registry removes old files across versions
  • Hash Registry - Tracks file states in .dino/file-hashes.json for smart detection

• New /dino.init Skill - In-session update command

  • Run dino init from within Claude Code sessions
  • Reports created, updated, and removed files
  • Safe to run repeatedly after package upgrades
  • Alternative to --force which destroys session context

• Deprecated Files Module - Centralized deprecation management

  • src/core/updater/deprecated-files.ts with versioned registry
  • 25+ deprecated files tracked across versions (v1.x → v16.x)
  • Auto-cleanup on every dino init run
  • cleanupDeprecatedFiles(), findExistingDeprecatedFiles() APIs

• Benefits

  • Zero Context Loss - Session state preserved across updates
  • Customization Safe - User changes detected and protected
  • Cleaner Projects - Deprecated files automatically removed
  • Better UX - Clear reporting of what changed

v16.1.0 - Phase 4: Enhanced Detection & Loading (Tier 3)

• Advanced Demand Signal Detection - ML-style semantic signal analysis for auto-injection

  • Synonym Map Engine (src/core/context/demand-signal-patterns.ts)
    • 8 signal roots with 41+ synonyms (file, function, class, error, test, memory, code, debug)
    • Multi-line detection across adjacent lines
    • Conditional patterns (if X then Y)
    • Temporal patterns (before/after X do Y)
    • Scope-aware patterns (within X, find Y)
  • Semantic Signal Detector (src/core/context/semantic-signal-detector.ts)
    • Advanced pattern matching with confidence scoring
    • Context-aware signal refinement
    • Feature flag: advancedDemandSignals (default: true)
  • 6 New Tests - Demand signal pattern coverage

• REPL Reconnaissance - Smart peek-before-load with confidence scoring

  • Confidence Scorer (src/core/context-repl/confidence-scorer.ts)
    • Formula: confidence = (richness * 0.4) + (recency * 0.3) + (relevance * 0.3)
    • Multi-dimensional evaluation: richness, recency, relevance
  • Reconnaissance Engine (src/core/context-repl/reconnaissance-engine.ts)
    • Strategy selection: breadth-first, depth-first, frequency-based
    • Peek optimization reduces unnecessary full loads by 30-40%
    • Feature flag: replReconnaissance (default: true)
  • 130 New Tests - Comprehensive REPL confidence coverage

• Smart Prefetch Intelligence - Transitive dependency analysis with learning

  • Smart Prefetch Engine (src/core/context/smart-prefetch-engine.ts)
    • Transitive dependency scoring: direct (0.85), dependent (0.75), depth-1 (0.7), depth-2 (0.5)
    • Graph traversal with cycle detection
    • Budget-aware prefetch limits
  • Prefetch Learner (src/core/context/prefetch-learner.ts)
    • Outcome tracking: hit rate = prefetchedAndUsed / totalPrefetched
    • Learning threshold: 50 outcomes required
    • Baseline target: 50%, Optimized target: 65%
    • Persistent learning in .dino/metrics/prefetch-outcomes.json
    • Feature flags: smartPrefetch, prefetchLearning (both default: true)
  • 62 New Tests - Prefetch intelligence coverage

• Focus Tag Expansion - Proximity-aware tool filtering with dynamic learning

  • Proximity Tags (src/mcp/proximity-tags.ts)
    • Three proximity levels: in-focus, near-focus, out-of-focus
    • Path overlap calculation for proximity classification
    • Tool tags include proximity for granular filtering
  • Focus Tag Learner (src/mcp/focus-tag-learner.ts)
    • Dynamic learning from tool usage patterns
    • Associates tools with focus areas based on actual use
    • Improves filtering accuracy over time
    • Feature flags: focusProximityTags, dynamicFocusTagLearning (both default: true)
  • 60+ New Tests - Focus tag expansion coverage

• Token Savings Enhancement - +5-10% additional savings from Phase 4 features

  • Advanced demand signals reduce false positives in auto-injection
  • REPL reconnaissance avoids 30-40% of unnecessary full loads
  • Smart prefetch reduces manual file loading overhead
  • Focus tag expansion improves tool filtering precision
  • New Total: 75-95% context savings (up from 70-90%)

v15.0.0 - Session-Based Memory Decay

• Session Memory Decay - Aggressive aging of previous session content
  • Memory Decay Engine (src/core/context/memory-decay.ts)
    • Ages content from previous sessions, keeps current session prominent
    • Items marked with [age:N] tracking session age
    • Items beyond threshold (default: 3 sessions) moved to Archive section
    • Configurable decay factor (default: 0.5 = 50% decay per session)
  • MCP Tool - dino_memory_decay for manual/automatic execution
    • Wired to session-start hook with automatic execution
    • Uses conversation ID or timestamp as session identifier
    • Respects sessionMemoryDecay feature flag (default: true)
  • Strategic Session Positioning - Sections ordered for LLM attention
    • Critical info (Phase, Blockers, Focus) at top
    • Recent context (Changes, Test Status) in middle
    • Historical content (Archive) at bottom
    • Based on "Lost in the Middle" research
  • Prefetch Integration - Auto-prefetch wired to post-edit hook
    • Executes dino_prefetch on focus changes
    • Respects budget < 70% threshold
    • Reduces manual file loading
  • 16 New Tests - Memory decay engine coverage
  • 54 Integration Tests - Hook→engine wiring verification

v14.1.0 - Subagent-First Operations

• Comprehensive Subagent Architecture - Autonomous code review, research, and quality assurance
  • Ambiguity Resolver - Auto-research when ambiguity > 40
    • Spawns research subagents with specialized context
    • Caches findings in .dino/research/
    • Environment variable: DINO_AMBIGUITY_THRESHOLD (default: 40)
  • Confidence Gate - Quality threshold enforcement
    • Gates at 80% (standard) or 90% (Ralph-eligible)
    • Triggers research or clarification automatically
    • Max 2 research rounds, 5 clarification rounds
    • Environment variables: DINO_CONFIDENCE_THRESHOLD, DINO_RALPH_CONFIDENCE_THRESHOLD
  • Reviewer Loop - Mandatory code review for Edit/Write
    • Security: hardcoded secrets, eval, innerHTML, SQL injection
    • Quality: style, performance, test coverage
    • Blocks on critical findings with DINO_MANDATORY_REVIEW=true
  • Feedback Collector - Structured quality tracking
    • Quality scoring (0-100) with trend analysis
    • Plateau detection (<2% improvement over 3 iterations)
    • Early stopping on max iterations (default: 10)
    • 5 MCP tools: dino_feedback_collect, _status, _learnings, _resolve, _reset
  • Skill Auditor - Pre-validation for planning/implementation
    • Audits /dino.nest, /dino.hunt, /dino.ralph before execution
    • PRD quality: stories, acceptance criteria, test commands, files
    • Ralph requires 90 score vs 80 for standard
  • Parallel Reviewer - Concurrent aspect reviews
    • Runs style/security/tests in parallel
    • Budget: 40% style, 30% security, 30% tests
    • Merges into unified review summary
    • Environment variable: DINO_PARALLEL_REVIEW (default: false)
  • Research Agent - Read-only exploration agent
    • Structured JSON output with findings/confidence
    • Disallows Edit/Write/Bash for safety
  • 128 New Tests - Comprehensive coverage across 6 agent modules
  • Auditor Pre-Checks - /dino.nest and /dino.hunt now validate before execution

v13.6.0 - RLM-Style Context REPL Environment

• Context REPL - Programmatic context access inspired by Recursive Language Models
  • REPLEnvironment - Core interface for context operations
    • execute() - Run queries with 2000-token result truncation
    • peek() - Zero-cost reconnaissance of context structure
    • grep() - Pattern matching with relevance scoring
    • partition() - Segment by directory/feature/dependency/time
  • Context Window Monitor - Track and control context growth
    • Growth tracking per operation (tokens before/after)
    • Growth rate calculation (tokens/minute)
    • Threshold notifications: 50% info, 75% warning, 90% critical
    • Export growth history for debugging
  • MCP REPL Tools - Three new tools for programmatic access
    • dino_repl_peek - All phases (reconnaissance)
    • dino_repl_grep - All phases (search)
    • dino_repl_execute - Implementation/testing only
    • Rate limiting: 10 calls/minute, 30s timeout
  • Auto-Inject Integration - REPL peek used for smarter context loading
    • Peek before loading for relevance filtering
    • Graceful fallback when REPL disabled
  • Token Savings - 30-40% additional savings from REPL features
    • Peek before load: 10-20%
    • Result truncation: 15-25%
    • Auto-inject integration: 5-10%
  • Combined Savings - 80-100% total context savings (with all engines)
  • 88 New Tests - Comprehensive REPL coverage, 100% passing
  • Feature Flag - enableReplEnvironment (default: false, opt-in)

v13.5.0 - Context Engineering Engines Operational

• Context Engineering Hook Wiring - All orphaned engines now wired and operational
  • Auto-Injection - Context automatically injected based on prompt demand signals
    • New MCP tool: dino_auto_inject - Manual context injection trigger
    • Wired to user-prompt-submit hook for automatic execution
    • Detects memory, decision, pattern, and failure queries
    • Respects autoContextInjection feature flag
  • Health Monitoring - Context health checked on session start
    • New MCP tool: dino_health_check - Get context health report
    • Reports rot index (0-100), health status, recommendations
    • Includes v13.4.0 focus fragmentation metrics
    • Outputs warnings when rot > 40%, critical when rot > 70%
    • Respects contextHealthMonitor feature flag
  • Auto-Unload - Cold resources automatically evicted at budget thresholds
    • New MCP tool: dino_auto_unload - Manual resource unload trigger
    • Wired to session-start at 70%/80%/90% budget thresholds
    • Light/medium/aggressive aggressiveness based on budget pressure
    • Evicts files not accessed in 10+ min (cold) or 30+ min (stale)
    • Respects autoUnload feature flag
  • Auto-Prefetch Suggestions - Focus changes trigger prefetch recommendations
    • Detects focus changes from edit clustering
    • Suggests @dino://focus resource when focus shifts and budget < 70%
    • Integrated into post-edit hook
  • Combined Impact - 70-95% total context savings (up from 60-85%)
  • 124 New Tests - Comprehensive integration and tool coverage

v13.4.0 - Focus-Aware Context Engineering

• Focus-Based Context Optimization - Context management adapts to current focus area
  • Focus-Aware Tool Filtering - Only load MCP tools relevant to focus area (~10-15% additional savings)
    • Wildcard tag system (['*']) ensures critical tools always available
    • Focus tags: core, mcp, commands, memory, ralph, agents, analysis, test, rules, skills
    • Combines with phase filtering for maximum efficiency
  • Focus-Aware Resource Deferring - Lower token thresholds for out-of-focus files (~5-10% additional savings)
    • In-focus: 500 tokens, Near-focus: 400 tokens, Out-of-focus: 300 tokens
    • Proximity calculated from path overlap with current focus
    • Automatically adjusts on focus changes
  • Focus-Aware Health Monitoring - Track context fragmentation relative to focus
    • Focus fragmentation score (0-100) measures context scatter
    • Focus distribution (in-focus/near-focus/out-of-focus file counts)
    • Recommendations trigger at >50% and >70% fragmentation
  • Combined Savings - 70-80% total context savings (up from 50-70%)
  • Feature Flags - 2 opt-out flags (focusFiltering, focusAwareDefer)

v13.3.0 - Full Auto Context Engineering

• Automatic Context Management - Zero manual CLI intervention for context engineering
  • Auto Context Injection - JIT loads files/symbols based on demand signals from prompts
  • Auto Context Unload - Evicts cold (>10 min) and stale (>30 min) resources at 70%/80%/90% budget pressure
  • Auto Prefetch - Preloads related files on focus change using dependency graph and access patterns
  • Context Health Monitor - Calculates rot index, provides health scores and recommendations
  • Memory Auto-Recall - Detects memory keywords (decided, why, previous) and auto-injects top 3 relevant memories
  • Feature Flags - 6 opt-out flags (autoContextInjection, autoPrefetch, autoUnload, autoStagedCompact, contextHealthMonitor, autoMemoryRecall)

v13.0.0 - Performance & Intelligence Suite

• Performance Optimizations
  • Phase detection caching with 5s TTL
  • Incremental dependency graph updates (O(changed) vs O(n))
  • Focus detection caching with directory patterns
• Context Engineering
  • Auto-lazy context loading with access tracking
  • Context pressure alerts at 50/70/90% thresholds
  • Test impact analysis for changed files
• Memory System
  • Auto-memory consolidation (>100 memories trigger)
  • Memory access heatmap tracking
  • Cross-session global memories
• Developer Experience
  • Circuit breaker dashboard HUD
  • Predictive tool recommendation engine
  • Auto-generated MCP tool documentation
• Architecture
  • Smart pre-validation v2 with pattern trust
  • Memory module split (store/retrieval/lifecycle)
  • Type safety audit with strict response builders
• New Capabilities
  • Context checkpoint/restore system
  • Semantic diff with change intent detection
  • Agent communication protocol (message passing + blackboard)

Core Features

• MCP Integration - 74 tools + 29 resources auto-configured during init
• Ralph Loop (v11.6.0) - Fresh-context iterative execution for large-scale development
  • Story-Based PRD - Define multi-story features with acceptance criteria
  • Fresh Context - Each iteration gets clean context (no history accumulation)
  • Git Integration - Full checkpoint/restore workflow with automatic commits
  • Learning Consolidation - Multi-source learning aggregation from commits, tests, and user input
  • PRD Generation - Structured product requirement document creation with story validation
  • Task Integration - Seamless interaction with dino task system
  • Metrics Tracking - Iteration success rates, checkpoint lifecycle, and story progress
  • Max 10 iterations - Hard limit with configurable max attempts per story
  • MCP tools: dino_ralph_init, dino_ralph_iterate, dino_ralph_status, dino_ralph_learn, dino_ralph_restore, dino_ralph_cleanup
• MCP Tool Aggressive Filtering (v10.3.0) - Phase-based tool context reduction (~50-70% savings)
  • Phase Filtering - Only load tools relevant to current workflow phase
  • Description Modes - Minimal/standard/full verbosity control
  • Lazy Details - Full descriptions loaded on-demand via dino://tools/details
  • Auto-Enabled - Filtering enabled by default, opt-out with DINO_AGGRESSIVE_FILTERING=false
• Lazy Context Loading (v10.1.0) - Defer file loading to save tokens (~60-80% initial context reduction)
  • Deferred Resources - Replace file content with lightweight placeholders (~5-10 tokens)
  • On-Demand Expansion - Load content only when needed with dino_expand
  • Token Budget - Control expansion with budget constraints
  • HUD Integration - Real-time display of deferred/loaded resources
• Self-Validate Hooks (v9.16.0) - Proactive validation system that catches edge cases before execution
  • Ambiguity Detection - Identifies vague terms, incomplete requirements, missing specifics
  • Edge Case Analysis - Null checks, empty collections, error handling coverage
  • Confidence Scoring - Weighted validation with thresholds (block < 0.4, ask < 0.6, proceed ≥ 0.6)
  • Impact Assessment - Breaking changes, security risks, cross-cutting concerns
  • Question Generation - Auto-converts issues to user questions with recommended options
  • Pre-validate hook blocks low-confidence operations before execution
• Phase 4 Advanced Optimizations (v9.14.0)
  • MCP Tool Filtering - Phase-based tool tiering for 60-70% context overhead reduction
  • Memory Recitation - SM-2 spaced repetition for long-term memory retention
  • Entropy Enhancement - Language-specific profiles and cross-section deduplication
  • External Vector DB - Qdrant/Pinecone adapter support for scalable embeddings
• Enhanced Observability - Execution tree, memory access log, context transitions with HTML/JSON export (v9.10.0)
• Session State - Automatic tracking of focus, test/build/lint status, blockers
• Persistent Memory - Decisions, patterns, learnings survive across sessions
• Interactive Memory Browser - Full-featured TUI for memory management with search, graph view, and actions (v10.0.0)
• Real-Time HUD - Statusline integration showing budget, phase, tests, focus, git status (v9.1.0)
• Session Handoff - Context continuity system for seamless session transitions (v9.1.0)
• Workflow Skills - Phases from discovery to completion (/dino.*)
• Optimized Test Suite (v12.3.0) - 74% performance improvement through intelligent parallelization
  • Two-Tier Workspace - Unit tests run in parallel (threads), integration tests sequential (forks)
  • Template Repo Pattern - Git repos copied from pre-initialized template (40% faster git-heavy tests)
  • Direct Imports - Function-level testing instead of CLI spawning (99% faster for session tests)
  • Test Commands - test:unit (< 30s), test:integration (< 2min), full suite (< 3min)
  • Performance - 2,857 tests in 2.5 minutes (was 9.5 minutes)
• Smart Test Scoping (v11.5.0) - /dino.release intelligently analyzes change impact to optimize test runs
  • Change Impact Analysis - Detects dependency/structural/config changes requiring full test suite
  • Targeted Testing - Isolated changes run focused tests via --testPathPattern
  • Skip When Verified - Avoids redundant tests when already passed < 10 min ago
  • Import Analysis - Greps for widespread dependencies to recommend full vs targeted scope
• Task Delegation - Specialized agents with context handoffs
• Recursive Agents - Self-spawning agents with complexity-based depth control (v8.0.0)
• Recursive Retrieval (v10.4.0) - Enhanced RLM-style query decomposition with:
  • Semantic Decomposition - Intent-based query splitting by distinct concepts
  • Dependency-Aware Decomposition - Sequential sub-queries with prerequisite analysis
  • Composition Decomposition - Parallel execution for "A and B" style queries
  • Quality Confidence Bounds - Uncertainty tracking with Wilson score intervals
  • Task-Specific Weight Calibration - Automatic quality weight adjustment per task type
  • Provenance Tracking - Full source chain attribution for every result
  • Conflict Detection - Automatic identification of score variance and data mismatches
  • Decision Rationale - Explainability with alternatives considered
• Dynamic Partitioning - Intelligent codebase segmentation (directory, feature, dependency, time)
• Context REPL - Declarative DSL for model-driven exploration
• Spec-Kit: Intelligent Prompt Analysis (v10.5.0 - v10.8.0) - Auto-specification system that analyzes prompts before implementation
  • Complexity Scoring - Ambiguity, scope, risk, novelty analysis (0-100 scale)
  • Confidence Gating - 80% confidence threshold prevents premature implementation
  • Auto-Trigger - Automatic /dino.spec suggestion for complex prompts (3+ signals)
  • Multi-Part Decomposition - Splits "add X and Y" into separate tasks
  • Context-Aware Questions - References existing patterns and focus area
  • Learning System - Refines scoring weights based on outcome feedback
  • Confidence Visualization (v10.8.0) - Progress bar, factor breakdown, improvement suggestions
  • MCP tools: dino_spec_analyze, dino_spec_feedback, dino_spec_stats, dino_spec_weights
• Context Engineering - Research-backed token optimization (v9.6.0, v9.7.0, v9.14.0)
  • Tool result summarization (~50-70% better retention)
  • Positional context optimization (~15-25% recall improvement)
  • Just-in-time context loading (~60-80% initial context reduction)
  • Entropy-guided budgeting (~20-30% token reduction)
  • Virtual memory management (~40-50% memory overhead reduction)
  • Multi-agent context isolation (~70-80% token reduction for parallel agents)
  • Semantic memory validation (~20-30% better relevance)
  • Language-specific entropy profiles (v9.14.0)
  • Cross-section deduplication (v9.14.0)
  • Reference docs and skills (/research, /context-optimize)

Installation

Standalone Binary (Recommended)

One-line install - no dependencies required:

macOS/Linux:

gh api repos/xullul/dino/contents/install.sh -H "Accept: application/vnd.github.raw" | bash

Windows PowerShell:

gh api repos/xullul/dino/contents/install.ps1 -H "Accept: application/vnd.github.raw" | iex

Note: Binaries are built via /dino.release workflow. Run dino release to publish new versions.

Building Binaries

bun build-binary.js                           # Build for all platforms
bun build-binary.js --target bun-windows-x64  # Windows only

Package Managers

With Bun:

bun add -g dino

With npm (alternative):

bun add -g dino

Quick Start

cd your-project
dino init          # Analyze codebase, install hooks, configure MCP

# MCP server is auto-configured for Claude Code
# Use /mcp in Claude Code to verify

CLI Commands

| Command | Description | |---------|-------------| | dino init | Initialize project (hooks + MCP) | | dino init --minimal | Minimal CLAUDE.md with MCP resources | | dino init --bootstrap | Include project analysis summary | | dino init --validate | Validate Claude Code configuration (v17.1.0) | | dino setup | Interactive wizard for provider configuration (v11.5.0, OAuth fix v11.5.1) | | dino setup --provider <name> --key <key> | Non-interactive provider setup (v11.5.0) | | dino provider list | List configured providers (v11.4.0) | | dino provider setup | Generate shell functions for providers (v11.4.0) | | dino status | Show session state and health | | dino refresh | Re-analyze codebase | | dino update | Update everything (skills, rules, hooks, MCP) (v10.9.0) | | dino update --scope mcp | Fix MCP configuration | | dino memory list | Show persistent memories | | dino memory search <query> | Search memories | | dino memory browse | Interactive memory browser TUI (v10.0.0) | | dino retrieve <query> | Hybrid file retrieval | | dino hud install | Install statusline integration (v9.1.0) | | dino hud preview | Preview HUD output (v9.1.0) | | dino hud debug | Debug context tracking data (v10.10.0) | | dino hud debug-clear | Clear HUD debug logs (v10.10.0) | | dino handoff create | Save session context for continuity (v9.1.0) | | dino handoff load | Load handoff context (v9.1.0) | | dino mcp serve | Start MCP server (auto-managed) | | dino mcp health | Check MCP health and diagnose issues |

Skill Store

Installable skills for common frameworks and tools. Install via dino store install <skill>:

Available Skills (v17.3.0)

Database & ORM

• postgresql - PostgreSQL 18 optimization (AIO, indexing, vacuum tuning)
• prisma - Prisma ORM v6 (relations, migrations, type safety)

Frontend Components

• mui.datagrid - MUI X DataGrid v7 (prevents common AI mistakes)
• mui.grid - MUI Grid layout v6+ (prevents confusion with DataGrid)
• mui.datepicker-luxon - MUI DatePicker v7 + Luxon 3 (Thai Buddhist calendar)
• shadcn-radix - Shadcn/ui + Radix UI (composition, styling patterns)

Testing

• vitest - Vitest v2 (mocking, async testing, coverage)
• playwright - Playwright E2E (selector patterns, async handling)

Workflow

• dino - Dino-spec workflow phases (discover → fossil)
• dino.statusline - Status line display configuration

All skills include comprehensive "MOST COMMON AI MISTAKES" sections to prevent typical code generation errors.

Workflow Skills

Use in Claude Code with /dino.<phase>:

| Phase | Skill | Purpose | |-------|-------|---------| | Discover | /dino.discover | Capture requirements | | Scout | /dino.scout | Research codebase | | Sniff | /dino.sniff | Clarify ambiguities | | Nest | /dino.nest | Plan implementation | | Hatch | /dino.hatch | Break into tasks | | Hunt | /dino.hunt | Execute implementation | | Aging | /dino.aging | Verify with tests | | Fossil | /dino.fossil | Archive and commit |

Utilities: /dino.status (includes --next, --focus, --tips), /dino.handoff (v9.1.0), /dino.release (v9.4.0)

Spec-Kit: /dino.spec (v10.5.0) - Intelligent prompt analysis before implementation

Validation: /dino.validate-deps (v17.4.0) - Dependency security and quality validation

Research: /research (v9.7.0)

MCP Resources

Access via @dino:// in Claude Code:

| Resource | Description | |----------|-------------| | @dino://session | Focus, blockers, test status | | @dino://structure | Project architecture | | @dino://patterns | Coding conventions | | @dino://commands | Build/test/lint commands | | @dino://tasks | Task registry and progress | | @dino://handoff | Agent handoff context | | @dino://memories/summary | All memories overview | | @dino://retrieval/ranked | Session-relevant files | | @dino://traces/tree | Execution tree overview (v9.10.0) | | @dino://traces/memory-access | Memory access log (v9.10.0) | | @dino://traces/context-graph | Context transition graph (v9.10.0) | | @dino://traces/overview | Combined trace overview (v9.10.0) |

MCP Tools

Key tools available via Claude Code:

| Tool | Purpose | |------|---------| | dino_status | Get session state | | dino_focus_set | Set focus area | | dino_memory_recall | Semantic memory search | | dino_context_repl | Declarative context queries | | dino_recursive_query | RLM-style retrieval | | dino_partition | Intelligent codebase partitioning | | dino_agent_analyze | Analyze task complexity for spawn decisions | | dino_agent_spawn_decision | Get recursive agent spawn recommendation | | dino_agent_tree | Display agent execution tree | | dino_task_create | Create task in registry | | dino_plan_create | Create implementation plan | | dino_graph_query | Query knowledge graph entities | | dino_graph_relations | Find entity relationships | | dino_graph_architecture | Architectural analysis (cycles, dependencies) | | dino_retrieve | Hybrid code retrieval (keyword + semantic + graph) | | dino_retrieve_similar | Find similar files | | dino_embed_search | Semantic code search with embeddings | | dino_embed_stats | Vector store statistics | | dino_chunk_file | AST-based semantic chunking | | dino_chunk_analyze | Chunking analysis | | dino_effort_recommend | Recommend effort level based on task complexity (v9.9.0) | | dino_trace_export | Export execution traces to JSON or HTML (v9.10.0) | | dino_memory_access_stats | View memory operation statistics (v9.10.0) | | dino_context_transitions | View context transition history (v9.10.0) | | dino_expand | Expand deferred context references on demand (v10.1.0) | | dino_deferred_status | Get status of deferred resources and lazy loading stats (v10.1.0) | | dino_defer | Defer a file for lazy loading instead of loading immediately (v10.1.0) | | dino_repl_execute | Execute REPL queries with result truncation (v13.6.0) | | dino_repl_peek | Zero-cost reconnaissance of context structure (v13.6.0) | | dino_repl_grep | Pattern matching within loaded context (v13.6.0) |

Specialized Agents

| Agent | Role | Purpose | |-------|------|---------| | @dino-analyst | Explorer | Read-only codebase research | | @dino-architect | Planner | Design and planning | | @dino-implementer | Implementer | Full implementation | | @dino-tester | Tester | Test writing/execution | | @dino-reviewer | Reviewer | Code review and QA |

Memory System

Semantic memory retrieval via MCP tools:

dino memory list              # View all memories
dino memory search "auth"     # Search memories

Project Structure

your-project/
├── CLAUDE.md                 # Project context
├── .claude/
│   ├── skills/dino/          # Workflow skills
│   ├── rules/dino/           # Workflow rules
│   └── agents/               # Specialized agents
└── .dino/
    ├── session.md            # Session state
    ├── structure.json        # Codebase analysis
    ├── tasks/                # Task registry
    ├── checkpoints/          # Agent checkpoints
    └── archive/              # Archived handoffs/research

Requirements

• Bun >= 1.0.0 (recommended) or Node.js >= 20.0.0
• Claude Code CLI

Troubleshooting

MCP not showing in /mcp

# Windows
claude mcp add-json --scope user dino '{"command":"cmd","args":["/c","bunx","dino","mcp","serve"]}'

# macOS/Linux
claude mcp add --scope user dino -- bunx dino mcp serve

Or use the direct command (if dino is in PATH):

claude mcp add --scope user dino -- dino mcp serve

Then restart Claude Code.

License

MIT

Built for developers who believe AI coding should be structured, reproducible, and context-aware.