ce-work-integration

Adds Task Bundle persistence and failure state machine tracking to ce:work execution.

6|2|Updated Jan 30, 2026
One-click install
npx skills add https://github.com/Jerrylalala/compound-engineering --skill ce-work-integration-jerrylalala
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ce-work-integration
Source: https://github.com/Jerrylalala/compound-engineering/tree/main/plugins/compound-engineering/skills-custom/ce-work-integration
Command: npx skills add https://github.com/Jerrylalala/compound-engineering --skill ce-work-integration-jerrylalala

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Long-running or multi-agent tasks lose context when interrupted, and blocked work has no recorded state, forcing you to restart from scratch or guess where execution stopped. ## Core Features & Use Cases - Task Bundle Persistence: Reads and writes state.md files under docs/tasks/<task-id>/ to track current step, status, blockers, and checkpoints across sessions. - Failure FSM: Converts implicit STOP behavior into an explicit state machine (active, blocked, debugging, replanned, resumed, reviewed, compounded) with a mandatory user confirmation gate before resuming a replanned task. - Ralph-lite Repair Loop: Runs a bounded two-attempt fix cycle against verifiers (tests, lint, build) before escalating to a blocked state. - Use Case: You resume a multi-step feature task after a CI failure; the skill loads state.md, shows the failure history, and continues from the exact step where work stopped instead of re-executing completed steps. ## Quick Start Run ce:work on a task that has a docs/tasks/<task-id>/ directory and let this overlay load state.md to resume from the last checkpoint.

Frequently Asked Questions about ce-work-integration

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I resume an interrupted task without losing progress?▼

Keep a Task Bundle at docs/tasks/<task-id>/ with a state.md file recording current_step and status. When ce:work runs, this overlay reads state.md and continues from the last checkpoint instead of re-executing completed steps.

How do I track blocked tasks in an AI coding workflow?▼

Use the explicit failure FSM: blockers trigger an active-to-blocked transition with a blocked_by reason (dependency, CI failure, unclear requirement, tool unavailable, or human decision) written into state.md along with a timestamped state history table.

When should I create a Task Bundle for a task?▼

Create one when the task has a docs/tasks/<task-id>/ directory, is expected to take over an hour, or involves multiple sub-agents. Short single-step tasks do not need the persistence overhead.

Does the task resume automatically after a plan change?▼

No. The replanned-to-resumed transition has a mandatory confirmation gate: the new plan summary is shown and the user must explicitly confirm before status changes back to active and execution restarts.

What happens when tests keep failing during task execution?▼

The Ralph-lite repair loop runs at most two fix attempts, each requiring a changed hypothesis or strategy. If verification still fails, the task transitions to blocked with the failing test recorded in state.md.