skills/ralph-tasks/SKILL.md
This skill should be used when the user asks to "add a task for Ralph", "list Ralph tasks", "remove Ralph task", "setup Ralph tasks", or mentions ralph-tasks, task queue for AI agent, or autonomous task management. Manages a JSON-based task queue for Ralph autonomous loop.
npx skillsauth add devattom/.claude ralph-tasksInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
<quick_start>
/ralph-tasks setup # Initialize in project
/ralph-tasks add Add authentication # Add task (quotes optional)
/ralph-tasks add Add dark mode -s # Add with context search
/ralph-tasks add Fix bug --search # Same with --search
/ralph-tasks list # List all tasks
/ralph-tasks remove 1 # Remove task #1
/ralph-tasks wake-up # Show loop command
Multi-task: Add multiple tasks in one message - they're processed in parallel. </quick_start>
<commands> | Command | Description | |---------|-------------| | `setup` | Initialize .claude/ralph-tasks/ with all files | | `add <task>` | Add a task (quotes optional) | | `add <task> -s` | Add with codebase context search | | `list` | Show all tasks with status | | `remove <id>` | Remove task by ID | | `wake-up` | Show command to start loop |Aliases: add tasks, add task, list tasks, remove task
</commands>
<file_structure>
{project}/.claude/ralph-tasks/
├── tasks.json # Task queue
├── wake-up.sh # Main loop script
├── prompt.md # Agent instructions
└── progress.txt # Accumulated learnings
</file_structure>
<entry_point>
FIRST ACTION: Load steps/action-init.md
This step parses the command and routes to the appropriate action file. </entry_point>
<action_files>
| Action | File | Purpose |
|--------|------|---------|
| init | steps/action-init.md | Parse command and route |
| setup | steps/action-setup.md | Initialize ralph-tasks |
| add | steps/action-add.md | Add task to queue |
| add-search | steps/action-add-search.md | Add task with context |
| list | steps/action-list.md | Display tasks |
| remove | steps/action-remove.md | Remove task by ID |
| wake-up | steps/action-wake-up.md | Show loop instructions |
</action_files>
<state_variables>
| Variable | Type | Description |
|----------|------|-------------|
| {command} | string | Parsed command (setup, add, list, remove, wake-up) |
| {task_description} | string | Task description (for add) |
| {task_id} | number | Task ID (for remove) |
| {search_mode} | boolean | Whether --search/-s flag is set |
| {project_path} | string | Current project path |
| {tasks_dir} | string | Path to .claude/ralph-tasks/ |
</state_variables>
<critical_rules> 🛑 NEVER run wake-up.sh automatically - user must run it themselves 🛑 NEVER modify tasks.json without proper JSON handling (use jq) ✅ ALWAYS use atomic file updates (write to tmp, then mv) ✅ ALWAYS load only the current action file (progressive disclosure) </critical_rules>
<success_criteria> ✅ Commands parsed correctly and routed to action files ✅ Each action executes its specific task ✅ Progressive loading minimizes context usage </success_criteria>
development
Use when you want to audit a project wiki for quality issues — stale version claims, contradictions between pages, orphan pages, broken wiki links, missing cross-references, or misalignment between wiki content and the actual codebase state.
development
Systematic error debugging with analysis, solution discovery, and verification
development
Structured adversarial debate between AI councillors using Agent Teams to evaluate ideas, plans, or decisions. ALWAYS use when the user says "council", "debate this", "evaluate this idea", "challenge my plan", "stress-test", "devil's advocate", "multiple perspectives", "évaluer cette idée", "débattre", "challenger mon plan", "tester cette décision", or when the user wants rigorous multi-perspective analysis of a proposal, architecture decision, or strategic choice. Each councillor (visionary, critic, pragmatist, innovator, ethicist, domain expert) represents a distinct perspective and they challenge each other through cross-examination and peer exchange, producing a nuanced verdict (PROCEED / PROCEED WITH CONDITIONS / RECONSIDER / DO NOT PROCEED). Do NOT use for divergent brainstorming or idea generation — use workflow-brainstorm instead.
testing
Automated CI/CD pipeline fixer - watches CI, fixes errors locally, commits, and loops until green. Use when CI is failing and you want to automatically fix and verify changes.