- Add Step 1.0 to test-gen.md to load source session task_description
- Add Step 1.0 to test-fix-gen.md (Session Mode) to load original user intent
- Include originalTaskDescription in session creation to enable semantic CLI selection
- Fixes data flow gap identified by Gemini review: user's CLI tool preferences
(e.g., "use Codex for fixes") now propagate through test workflow sessions
This ensures semantic CLI tool selection works correctly in derivative test sessions
by preserving the original user's task description from the source session.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Rename Track 0 from 'Exploration Aggregation' to 'Exploration Synthesis'
- Add explicit synthesis logic for critical_files prioritization and deduplication
- Enhance explore-json-schema to support relevance scores in relevant_files
- Update explore agent prompts across context-gather and lite-plan commands
- Add synthesizeCriticalFiles and synthesizeConflictIndicators helper functions
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- context-gather.md: Add Step 2 with complexity assessment (Low/Medium/High)
and parallel cli-explore-agent execution (1-4 agents based on complexity)
- context-search-agent.md: Add Track 0 for exploration aggregation with
exploration_results schema including all_patterns and all_integration_points
- conflict-resolution.md: Consume exploration_results for enhanced conflict
detection with pre-identified conflict indicators
- task-generate-agent.md: Use exploration critical_files for focus_paths
and all_patterns/all_integration_points for implementation guidance
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Revert cost-aware batching and multi-agent parallelism in lite-execute
- Remove execution_group and task complexity from schema
- Remove execution_group rules from lite-plan
- Keep 50k context threshold protection in lite-plan
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add executeBatch dispatcher to route batches to Agent/Codex
- Update executeBatchAgent to use batch.tasks instead of planObject.tasks
- Update executeBatchCodex to use batch.tasks instead of planObject.tasks
- Each parallel batch now dispatches to separate concurrent agent
Before: All batches → single agent with ALL tasks (no parallelism)
After: P1 → Agent 1, P2 → Agent 2 (true concurrent execution)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Schema loading is only needed for direct Claude planning (Low complexity).
The cli-lite-planning-agent handles schema internally for Medium/High complexity.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Updated execution flow diagram to show schema loading as REQUIRED step
- Added explicit schema reference for Low complexity direct planning
- Ensures plan.json follows consistent structure regardless of complexity level
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Dashboards use fetch() API which doesn't work with file:// protocol due to
browser CORS restrictions. Now starts Python HTTP server for proper access.
- review-module-cycle: port 8765
- review-session-cycle: port 8765
- review-fix: port 8766 (different port for concurrent use)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add errors array display with badge and detailed error list in Active Agents
- Enhance task cards with dimension badge, file:line, attempts, and commit hash
- Add Finding Details Modal with complete information on click
- Display fix_strategy and risk_assessment in Active Groups cards
- Add renderAgentErrors(), renderStrategyAndRisk() helper functions
- Increase JSON field consumption rate from ~53% to ~90%
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
This commit introduces an independent fix progress tracking system with
enhanced visualization and theme customization.
Changes:
- Create standalone fix-dashboard.html with dark/light theme toggle
- Add comprehensive task visualization with status filtering
- Implement real-time progress tracking with 3-second polling
- Add stage timeline with parallel/serial execution visualization
- Include flow control steps tracking for active agents
- Simplify state management (remove fix-state.json, use metadata in fix-plan.json)
- Update review-fix.md documentation with dashboard generation steps
- Change template references from @ to cat command syntax
Dashboard Features:
- Theme toggle (dark/light) with localStorage persistence
- Task cards grid with status-based filtering (all/pending/in-progress/fixed/failed)
- Tab-based view for Active Groups and Active Agents
- Enhanced progress bar with gradient and shimmer animation
- Stage timeline with visual status indicators
- Fix history drawer for session review
- Fully responsive design for mobile devices
Technical Improvements:
- Consumes new JSON structure (fix-plan.json with metadata field)
- Real-time aggregation from multiple fix-progress-{N}.json files
- Single HTML file (self-contained, no external dependencies)
- Static generation (created once in Phase 1, no subsequent modifications)
- Browser-side polling for status updates
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Problem: Export filename was 'fix-export-fix-export-1764148101709.json'
Root cause: export_id already contained 'fix-export-' prefix, then filename construction added it again
Solution: Keep export_id as timestamp only, let filename construction add the prefix
Result: Correct filename 'fix-export-1764148101709.json'
Move specific JSON structure requirements from cli-explore-agent (keep generic)
to review-*-cycle.md prompts. Key requirements now inline in prompts:
- Root must be array [{}] not object {}
- analysis_timestamp field (not timestamp/analyzed_at)
- Flat summary structure (not nested by_severity)
- Lowercase severity/id values
- Correct field names (snippet not code_snippet, impact not exploit_scenario)
