Releases: azalio/map-framework
Releases · azalio/map-framework
MAP Framework v3.2.0
[3.2.0] - 2026-02-14
Added
- Artifact-gated validation in MAP orchestrator for stricter workflow enforcement
- Enhanced skills with examples, troubleshooting sections, trigger rules, and validation scripts
- skip_step command for MAP orchestrator to allow controlled step skipping
Fixed
- Documentation accuracy audit (48 fixes): Comprehensive alignment of all docs, presentations, and templates with actual implementation
- Corrected agent count references across all docs (8/9/11 → 12 agents)
- Corrected command count references (updated to 10 MAP commands)
- Added missing agents (Synthesizer, DebateArbiter, ResearchAgent, FinalVerifier) to ARCHITECTURE.md and presentations
- Replaced phantom
/map-featureand/map-refactorreferences with implemented workflows - Removed stale haiku model references from presentations
- Fixed Evaluator workflow assignments and map-fast agent pipeline docs
- Template variable consistency: Resolved 8 template variable inconsistencies (
{{standards_url}}→{{standards_doc}}, etc.) - Branch sanitization: Unified branch name sanitization across all hooks, commands, and agents
- Path conventions: Corrected flat
.map/path references to nested<branch>/directory convention - API parameter naming: Fixed
top_k→limitin documentation-reviewer and other agents - MAP workflow inconsistencies: Resolved 35 audit issues across orchestrator, commands, and agent templates
- Plan path bug and evidence indentation in orchestrator
- Removed stale references: Cleaned up RETRY_LOOP/APPLY_CHANGES step references
- Test fixtures: Updated to cover all 12 agents and 10 commands
- Black formatting: Fixed formatting in 4 template/test files
MAP Framework v3.1.0
Changed (BREAKING)
- Hook-Based Context Injection: Optimize /map-efficient workflow with state-machine orchestration
- Problem: 995-line command file (5.4K tokens) caused attention dilution → 20% step compliance
- Solution: State-machine + PreToolUse hook injection → 85% predicted compliance
- Command file reduced: 995 → 394 lines (5.4K → 1.75K tokens, 68% reduction)
- New hook:
workflow-context-injector.py- Injects step reminders before every tool call - New state machine:
.map/scripts/map_orchestrator.py- Enforces 14-phase workflow sequencing - New utilities:
.map/scripts/map_step_runner.py- Deterministic step executors - State file:
.map/<branch>/step_state.json- Tracks current step phase for hook injection - Token efficiency: 54K → 9.25K per workflow (83% reduction despite hook overhead)
- Migration: Run
mapify initto update project structure with new hooks and scripts
- Simplified Workflow: Removed workflow-gate.py enforcement hook
- Actor now applies code directly with Edit/Write tools (no gate blocking)
- Monitor validates WRITTEN code by running tests, not proposals
- Simpler flow: Actor writes → Monitor tests → If issues, Actor fixes → Repeat
- Phase 2.7 renamed: APPLY_CHANGES → UPDATE_STATE (code already applied by Actor)
Added
- Ralph Wiggum Loop Integration: Continuous iteration pattern to prevent premature completion and hallucinated success
- State machine with 10 phases (INIT → DECOMPOSITION → EXECUTION → FINAL_VERIFICATION → COMPLETE/RE_DECOMPOSITION/ESCALATE/HARD_STOP/RECOVERY/WONT_DO)
- Circuit breaker with configurable limits (max 50 tool calls, 5 same-file edits, 60 min wall time)
- Final verification step in map-efficient.md (Step 3.5) with re-decomposition on failure
- Thrashing detection (oscillation detection via net_progress and confidence_variance)
- Recovery path via RESET_LIMITS marker file
- New Agent:
final-verifier.md- Adversarial verifier with Root Cause Analysis for Ralph Loop - New Hooks:
ralph-circuit-breaker.py(PreToolUse): Enforces iteration limits, blocks at thresholdsralph-iteration-logger.py(PostToolUse): Logs metrics, detects thrashing patternsralph-context-pruner.py(PreCompact): Archives old logs, truncates large files
- New Python Modules:
src/mapify_cli/ralph_state.py: State machine, circuit breaker config, verification types, thrashing detectionsrc/mapify_cli/dependency_graph.py: Cascade invalidation for subtask dependencies
- New Configuration:
.claude/ralph-loop-config.json- Single source of truth for Ralph Loop limits - New Reference:
.claude/references/escalation-matrix.md- Escalation decision rules
Changed
- task-decomposer.md: Enhanced with Acceptance Criteria table format, re-decomposition mode, dependency enforcement
- map-efficient.md: Added Step 3.5 Final Verification with circuit breaker check, final-verifier invocation, re-decomposition logic
- settings.hooks.json: Added PreToolUse, PostToolUse, and PreCompact hook entries for Ralph Loop
Documentation
- Branch-scoped artifacts stored in
.map/<sanitized-branch>/directory - Branch name sanitization (e.g.,
feature/foo→feature-foo) for safe filesystem paths
MAP Framework v3.0.0
[3.0.0] - 2026-01-16
Changed (BREAKING)
- Memory layer migration: Migrate from
playbook.db/cipher to mem0 MCP for all pattern storage. This is a breaking change that requires mem0 MCP server configuration.
Added
- P0 foundation implementation: security hooks, permissions system, workflow recovery
- mem0 MCP integration for tiered pattern storage (branch → project → org scopes)
- Project settings allowlist extensions for worktree, sourcecraft, mem0 MCP tools
Fixed
- Address PR #70 review feedback for P0 foundation
- Align documentation with actual implementation
- Workflow enforcement to prevent Actor→Monitor cycle skip
- Documentation fixes: ARCHITECTURE.md workflow diagrams, deprecated /map-feature /map-refactor references
- Code quality: Black formatting, ruff linting, mypy type errors
Documentation
- Complete migration of playbook.db/cipher references to mem0 MCP across all docs and templates
- Comprehensive documentation update to v2.3.0 standards
- README optimization (418→93 lines) for improved conversion
MAP Framework v2.3.0
[2.3.0] - 2026-01-10
Added
/map-planningskill: File-based planning for MAP Framework workflows with branch-scoped task tracking in.map/directory- Single-Writer Governance and 3-Strike Protocol for plan modification control
- Integration of map-planning skill with mapify templates and orchestrator
Fixed
- Critical bugs in map-planning skill session state management
Documentation
- Updated README and skills docs for map-planning skill
Installation:
pip install mapify-cli==2.3.0Full Changelog: https://github.com/azalio/map-framework/blob/main/CHANGELOG.md
MAP Framework v2.2.0
[2.2.0] - 2026-01-08
Added
/map-debatecommand: Debate-based MAP workflow with Opus arbiter for multi-variant synthesis. Generates 3 Actor variants in parallel (security/performance/simplicity focus), validates with parallel Monitors, then usesdebate-arbiter(Opus model) to cross-evaluate and synthesize optimal solution
Changed
- Documentation cleanup: Remove deprecated
/map-featurereferences, update learning workflow info
Fixed
- Address reviewer feedback on map-debate documentation
Installation:
pip install mapify-cli==2.2.0Full Changelog: https://github.com/azalio/map-framework/blob/main/CHANGELOG.md
MAP Framework v2.1.0
Added
- External static analysis scripts for Monitor agent (
analyze.sh,lint-go.sh,lint-python.sh) - LLM Council recommended improvements to MAP workflow (context7 integration, parallel execution)
Changed
- Optimize task-decomposer template with references to mapify init
- Extract common functions to shared module with tests
- Update README and sync templates with map-efficient improvements
Fixed
- Security hardening per Copilot review
- Improve clarity per Copilot review comments (multiple rounds)
- Fix agent count documentation (8→10) and update template sync
- Fix black formatting issues
Documentation
- Document map-efficient command template
- Sync map-efficient.md documentation with source template
MAP Framework v2.0.0
[2.0.0] - 2025-12-15
Changed
- Parallelize Monitor, Predictor, Evaluator agents in
/map-reviewworkflow for improved performance - Auto-create
.mcp.jsonduringmapify initfor better MCP server integration
Fixed
- Remove hooks-related CI job and test after hooks system removal
- Restore JSON validation in stop.sh hook for malformed input handling
- Address Copilot and LLM Council security review findings
- Clarify enforcement points and framework-level secret handling in documentation
- Handle malformed JSON in stop.sh hook with updated INPUT FORMAT docs
- Address PR #56 review comments
- Fix black formatting issues
Added
- New research-agent for context isolation during research tasks
BREAKING CHANGES
Hooks System Removed
The Claude Code hooks system has been completely removed from MAP Framework.
Rationale:
- Hooks added complexity without proportional value
- Core MAP workflows (
/map-efficient,/map-debug,/map-fast) operate independently of hooks - Maintenance burden outweighed benefits
What was removed:
.claude/hooks/directory (13 hook scripts)src/mapify_cli/__init__.pyfunctions:load_settings_with_merge(),merge_hooks_settings(),install_hooks()src/mapify_cli/templates/hooks/directory- CLI option:
--with-hooks/--no-hooksfrommapify init - 59 test cases (test_hooks_*.py, test_init_merge.py, test_inject_playbook_bullets.py)
Migration guide:
For existing projects with hooks installed:
- Hooks are now user-managed - The
.claude/hooks/directory (if present) will be ignored by MAP Framework - No action required - Your existing hooks will continue to work as Claude Code hooks
- Optional cleanup - You can safely remove
.claude/hooks/if you don't use custom hooks
What continues to work:
- ✅ All MAP workflows (
/map-efficient,/map-debug,/map-fast,/map-learn,/map-release,/map-review) - ✅ Agent orchestration via Task tool
- ✅ Playbook management via
mapify playbookcommands - ✅ MCP server integration (cipher, context7, deepwiki, etc.)
What no longer works:
- ❌
mapify init --with-hooks/--no-hooksoptions (removed from CLI) - ❌ Automatic hooks installation via
mapify init - ❌ Hooks template synchronization
Upgrade path:
# Upgrade MAP Framework to v2.0.0
uv tool upgrade mapify-cli
# (Optional) Remove hooks directory if you don't use custom hooks
rm -rf .claude/hooks/[1.7.0] - 2025-12-08
[2.0.0] - 2025-10-17
Added - Agent Templates Overhaul
See Agent Templates CHANGELOG for complete v2.0.0 changes.
Summary:
- Comprehensive MCP integration framework
- XML-style semantic structure for better LLM parsing
- Template size: 2,232 → 9,269 lines (+258% for comprehensive guidance)
- Removed orchestrator as subagent (moved to slash commands)
For older changes and agent template details, see:
- Agent Templates CHANGELOG
- Git commit history
Versioning
Version Format: MAJOR.MINOR.PATCH
- MAJOR: Breaking changes (incompatible API/workflow changes)
- MINOR: New features (backward compatible additions like Phase 1)
- PATCH: Bug fixes and minor improvements
MAP Framework v1.7.0
[1.7.0] - 2025-12-08
Added
- Optional Learning Command: Added
/map-learncommand for optional post-workflow learning. Reflector and Curator agents are now invoked on-demand rather than automatically in workflows (cdc7e4e) - Auto-Approval Permissions:
mapify initnow configures auto-approval rules for common readonly operations (cipher memory search, tracker queries, sequential-thinking) to reduce permission prompts (18f9532)
Changed
- Workflow Simplification: Removed unused workflow commands (
/map-feature,/map-refactor) to reduce maintenance burden. Use/map-efficientfor feature work (cdc7e4e) - Permissions Merge: Settings permissions now use additive merge strategy to preserve user-defined rules (b585173, 1978af8)
Fixed
- Map-Review Command: Restored
/map-reviewcommand that was accidentally removed and updated stale agent references (1394935) - Stop Hook: Restored malformed JSON handling in
stop.shquality gates hook for robustness (41b96c9) - README Accuracy: Updated README to reflect actual available commands, fixed playbook bullet ID generation for consistent identifiers (af2d5d3)
- Documentation Consistency: Fixed Next Steps sections across commands to show actual available commands (c0a257d)
- Map-Learn References: Removed stale references to deleted commands in
/map-learntemplate (3fcf8fc) - Agent Instructions: Removed misleading 'orchestrator directly' instruction from agent templates (ea75b21)
- Type Safety: Resolved 39 mypy type errors across 11 files, improving code quality (fe474dd)
Removed
- Recitation Functionality: Removed
mapify recitationcommands and related functionality. This feature was underutilized and added maintenance complexity (a1be4f8) - MCP Server: codex-bridge: Removed codex-bridge MCP server from the framework (7a7e363)
- Removed from
INDIVIDUAL_MCP_SERVERSconstant - Removed from agent template generators (actor, predictor)
- Removed from
agent_mcp_mappingsconfiguration - Updated all agent templates to remove codex-bridge references
- Updated documentation (README, ARCHITECTURE, presentations)
- Updated
.mcp.json.exampleand plugin configuration - Updated tests to expect 5 MCP servers instead of 6
- Rationale: Simplify MCP server dependencies; codex-bridge functionality can be achieved through other tools
- Removed from
MAP Framework v1.6.2
[1.6.2] - 2025-11-29
Fixed
- MAP Efficient Workflow: Fixed incorrect
subagent_typeparameters in/map-efficientcommand template. Changed from deprecatedtypeparameter to correctsubagent_typefor all Task tool invocations (reflector, curator, monitor, predictor, evaluator) (e05793a)
Installation
pip install mapify-cli==1.6.2Full Changelog
https://github.com/azalio/map-framework/blob/main/CHANGELOG.md
MAP Framework v1.6.1
[1.6.1] - 2025-11-28
Fixed
- Playbook Migration: Fixed migration from
playbook.jsontoplaybook.dbwhen usingmapify init --force. The migration now properly detects and removes invalid/incompleteplaybook.dbfiles before attempting migration, and cleans up staleplaybook.jsonfiles after successful migration (7cfa82e) - Playbook References: Removed all
playbook.jsonreferences from codebase (except CHANGELOG history). Updated CLAUDE.md, agent templates, skills, and documentation to referenceplaybook.dbonly. Added clarifying comments to migration code and tests (fbe6bd3)
Installation
pip install mapify-cli==1.6.1