메뉴
Manages Ralph Wiggum Loop state files (TASKS.md, PROGRESS.md, LOOP_CONFIG.md). Use this skill when asked to initialize loop state, create task files, reset progress, compact progress logs, or manage the loop's persistent disk state.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | task-state-manager |
| description | Manages Ralph Wiggum Loop state files (TASKS.md, PROGRESS.md, LOOP_CONFIG.md). Use this skill when asked to initialize loop state, create task files, reset progress, compact progress logs, or manage the loop's persistent disk state. |
This skill manages the persistent state files used by the Ralph Wiggum Loop.
The loop uses three core state files, all in the project root:
| File | Purpose | Created by |
|---|---|---|
TASKS.md | Task list with status markers | User or loop-planner agent |
PROGRESS.md | Append-only iteration log | Loop orchestrator |
LOOP_CONFIG.md | Loop configuration | User or init-state.sh |
Run init-state.sh to create template state files:
bash .github/skills/task-state-manager/init-state.sh [--force]
Use --force to overwrite existing files.
Tasks in TASKS.md use these status markers:
| Marker | Meaning |
|---|---|
[ ] | Pending — not yet started |
[~] | In progress — being worked on |
[x] | Done — completed successfully |
[!] | Failed — attempted but unsuccessful |
[STOP] | Stop flag — halt the loop |
When marking a task complete:
### 1. Implement user model
- **Status:** [x]
- **Completed:** 2024-01-15T10:30:00Z
When marking a task failed:
### 3. Fix auth middleware
- **Status:** [!]
- **Error:** TypeError: Cannot read property 'token' of undefined
- **Attempted:** 2024-01-15T11:00:00Z
If PROGRESS.md grows too large (>500 lines), compact it:
To reset for a new loop run:
TASKS.md but reset all statuses to [ ]PROGRESS.md to PROGRESS.{timestamp}.mdLOOP_CONFIG.md as-isPROGRESS.md — it's the audit trailPROGRESS.md, never overwrite existing entriesTASKS.md is the source of truth for what work remainsLOOP_CONFIG.md controls safety limits — don't weaken them during a loopEnforces safety guardrails for the Ralph Wiggum Loop: iteration limits, path restrictions, diff size limits, and validation gates. Use this skill when checking if a loop operation is within safety bounds or when configuring loop safety parameters.
Runs the Ralph Wiggum Loop — an iterative fresh-context agent loop. Use this skill when asked to execute a loop, start iterating on tasks, or run the ralph wiggum pattern. Provides the loop execution script and instructions for invoking it.
Detects convergence issues in Ralph Wiggum Loop iterations: oscillation, stagnation, metric gaming, and context overflow. Use this skill when checking loop health or when the loop seems stuck.