- Created a new document for command and skill references, detailing orchestrator commands, workflow session commands, issue workflow commands, IDAW commands, with-file workflows, cycle workflows, CLI commands, memory commands, team skills, workflow skills, utility skills, and Codex capabilities.
- Added a comparison table for workflows, outlining their best uses, levels, self-containment, and automatic chaining behavior.
- Add explicit session termination output before lite-planex handoff to
prevent phase collision between analyze-with-file and lite-plan
- Restructure Phase 4: display conclusions summary before asking next step
- Simplify post-completion options to 3 (执行任务/产出Issue/完成),
remove redundant 导出报告 since report is always shown
- Update 01-lite-plan.md context isolation label to match new option name
- Updated phase references in SKILL.md and 01-lite-plan.md to use "LP-Phase" prefix for consistency.
- Added critical context isolation note in 01-lite-plan.md to clarify phase invocation rules.
- Enhanced execution process descriptions to reflect updated phase naming conventions.
Improve error handling in frontend routing
- Introduced ChunkErrorBoundary component to handle lazy-loaded chunk load failures.
- Wrapped lazy-loaded routes with error boundary and suspense for better user experience.
- Created PageSkeleton component for loading states in lazy-loaded routes.
Sanitize header values in notification routes
- Added regex validation for header values to prevent XSS attacks by allowing only printable ASCII characters.
Enhance mobile responsiveness in documentation styles
- Updated CSS breakpoints to use custom properties for better maintainability.
- Improved layout styles across various components to ensure consistent behavior on mobile devices.
Change CLI CONSTRAINTS from schema-driven JSON output to structured
markdown text output, matching the architecture where CLI produces text
plans and Claude agent parses them into JSON.
1. extractStructuredTasks regex: /### (TASK-\d+|T\d+):/ was too strict,
failing when Gemini outputs fewer #'s or omits colon. Relaxed to
/#{1,3}\s*(TASK-\d+|T\d+):?\s*/ to tolerate format variations.
2. CLI Command Template CONTEXT field: added explicit instruction to use
Prior Analysis from task description as primary planning context when
exploration files are absent.
Revert the Phase 2 skip when no explorations. CLI execution should always
run but use Prior Analysis block from task description as primary context
when exploration files are absent.
Global CLAUDE.md sets `run_in_background: true` as default for CLI calls,
which conflicts with the agent's need for synchronous results. The agent
may have followed the global default, firing CLI in background and getting
no output to parse. Strengthen the instruction to explicitly override the
global default with `run_in_background: false`.
When cli-lite-planning-agent receives empty exploration context (from
analyze-with-file → lite-plan flow), it previously spent all tool budget
on CLI execution (Phase 2) leaving no budget for file writing (Phase 4).
Add explicit "Prior Analysis Fast Path" in Execution Flow: when prompt
contains "No exploration files", skip Phase 2 and generate plan directly
from task description's Prior Analysis block, preserving tool budget for
mandatory file output (plan.json + .task/*.json).
- Deleted outdated templates for epics, product brief, and requirements PRD.
- Introduced lazy loading for locale messages in i18n module to enhance performance.
- Updated main application bootstrap to parallelize CSRF token fetching and locale loading.
- Implemented code splitting for router configuration to optimize bundle size and loading times.
- Added WebSocket connection limits and rate limiting to improve server performance and prevent abuse.
- Enhanced input validation with compiled regex patterns for better performance and maintainability.
- Rename skill directory from workflow-lite-plan to workflow-lite-planex (planex = plan + execute)
- Remove standalone lite-execute command entry from command.json and analyze_commands.py
- Update all 60+ files referencing workflow-lite-plan to use workflow-lite-planex
- Update descriptions to clarify Phase 1: plan → Phase 2: execute architecture
- Remove lite-execute as standalone command from orchestrator routing tables
- Update docs (EN/ZH) to reflect unified planex naming and phase descriptions
Add workflow boundary markers to analyze-with-file Post-Completion Options
so that when "生成任务" triggers lite-plan via Skill(), the model follows
lite-plan's Phase 1-5 exclusively instead of returning to analyze Phase 4
synthesis. Root cause: Skill() expands inline (shared context), causing
phase number collision between the two workflows.
- Fix McpManagerPage to read config after optimistic update and pass only expected fields
- Add debug logging for enabledTools config building and ccw-tools server saving
- Add dedicated greenfield (0→1) task type with complexity-adaptive routing
(low→brainstorm-to-plan, medium→greenfield-plan, high→greenfield-phased)
- Narrow roadmap regex to only /roadmap|路线.*图/ to prevent false matches
- Change brainstorm-to-plan chain from lite-plan to workflow-plan→execute
- Sync all changes between ccw.md and ccw-coordinator.md
fix(api): handle empty enabledTools array and improve default tool logic
fix(queueScheduler): ignore network errors in loadInitialState
fix(auth): ensure token generation handles max session capacity
chore(dependencies): update package requirements to use compatible version specifiers
chore(tests): add new test cases for incremental indexer and migrations
Add explicit constraints preventing Phase 3 interactive rounds from
modifying source code. Feedback handlers now clearly scope updates to
roadmap.md and issues.jsonl only, with code changes deferred to handoff.
Root cause: generateToken() returned undefined when session already
had maxTokensPerSession (5) tokens, causing ERR_HTTP_INVALID_HEADER_VALUE.
Fix: Force generate token even when at capacity, ensuring we always
return a valid token string.
Related: v7.1.1 CLI process hang fix
When analyze-with-file artifacts are passed to lite-plan, exploration is
skipped (hasPriorAnalysis=true) leaving manifest.explorations empty. The
agent prompt's exploration section rendered as blank, causing the planning
agent to lack context and fail to produce task JSON. Add conditional to
output a fallback hint directing the agent to use the Prior Analysis block
in the task description as primary planning context.
Root cause: HTTP Keep-Alive connections kept event loop alive,
preventing process.exit() from executing even after CLI_EXECUTION_COMPLETED
event was sent.
Fix: Add `agent: false` and `Connection: close` header to HTTP requests
in notifyDashboard() and broadcastStreamEvent() functions.
- agent: false - Creates new Agent per request instead of global Keep-Alive Agent
- Connection: close - Tells server to close connection after response
Fixes issue where `ccw cli --tool gemini` would complete execution but
Bash command would hang indefinitely.
- Remove HttpOnly from XSRF-TOKEN cookie for JavaScript readability
- Add hook installation status detection in system settings API
- Update InjectionControlTab to show installed hooks status
- Add brace expansion support in globToRegex utility
- Pre-task context analysis via gemini for bugfix/complex tasks
- CLI-assisted error diagnosis before retry on skill failure
- Consistent implementation across run.md and resume.md
- Added IDAW (Independent Development Autonomous Workflow) commands for batch task execution, including `/idaw:add`, `/idaw:run`, `/idaw:status`, and `/idaw:resume`.
- Updated documentation for IDAW commands in both English and Chinese.
- Modified favicon and logo SVGs to reflect new orbital design with dynamic colors.
- Incremented package version from 7.0.6 to 7.0.9.
- Updated `assembleSkillArgs` function in `resume.md` and `run.md` to sanitize task goal for shell safety by escaping special characters.
- Introduced a new animated orbital motion demo in `icon-concepts.html`, showcasing agents orbiting with varying speeds and a breathing core effect.
- Implemented /idaw:resume to resume interrupted sessions with task handling and auto mode.
- Created /idaw:run for executing task skill chains with git checkpoints and session management.
- Added /idaw:status for viewing task and session progress, including overview and specific session details.
- Introduced helper functions for task type inference and skill argument assembly.
- Enhanced task management with session tracking, progress reporting, and error handling.
- Introduced a detailed report outlining compliance issues and recommendations for the `ccw/frontend` implementation of Hook templates.
- Identified critical issues regarding command structure and input reading methods.
- Highlighted errors related to cross-platform compatibility of Bash scripts on Windows.
- Documented warnings regarding matcher formats and exit code usage.
- Provided a summary of supported trigger types and outlined missing triggers.
- Included a section on completed fixes and references to affected files for easier tracking.
- Fix path mismatch between command files and frontend/backend spec-index-builder
- Update init-specs.md, init-guidelines.md, sync.md, solidify.md to use .ccw/specs/
- Update init.md, start.md, clean.md, unified-execute-with-file.md, collaborative-plan-with-file.md
- Add scope field to architecture-constraints.md and coding-conventions.md
- Ensures specs created by commands are visible in frontend Spec Settings page
- Created a new document for custom skills development (`custom.md`) detailing the structure, creation, implementation, and best practices for developing custom CCW skills.
- Added an index document (`index.md`) summarizing all built-in skills, their categories, and usage examples.
- Introduced a reference guide (`reference.md`) providing a quick reference for all 33 built-in CCW skills, including triggers and purposes.
- Introduced comprehensive documentation for the queue management feature, detailing its pain points, core functionalities, and component structure.
- Added terminal dashboard documentation, highlighting its layout, core features, and usage examples.
- Created an index page in Chinese for Claude Code Workflow, summarizing its purpose and core features, along with quick links to installation and guides.
- team-worker: add fast_advance message bus log after spawning successor,
closing coordinator state blind spot during fast-advance
- team-worker: add Knowledge Transfer section with upstream loading,
downstream publishing, and context_accumulator conventions
- role-spec-template: add Knowledge Transfer Protocol with Transfer
Channels table and shared-memory.json namespaced write convention
- monitor.md (v2+v5): add fast-advance reconciliation step reading
fast_advance messages, add State Sync section for coordinator wake
- lifecycle-v5 SKILL.md: update cadence diagram with fast_advance log
