task-contract

Confirms acceptance criteria, scope, verification, and limits before starting open-ended tasks.

1|Updated Mar 12, 2025
One-click install
npx skills add https://github.com/void2610/dotfiles --skill task-contract-void2610
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: task-contract
Source: https://github.com/void2610/dotfiles/tree/main/.claude/skills/task-contract
Command: npx skills add https://github.com/void2610/dotfiles --skill task-contract-void2610

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Open-ended requests like "improve this as you see fit" or "handle it autonomously" lack explicit acceptance criteria, causing repeated correction loops after work is done. This Skill fixes the contract upfront in a single confirmation round before any work begins. ## Core Features & Use Cases - Four-Point Contract Confirmation: Asks exactly once about acceptance criteria, editable scope, verification method with reporting cadence, and iteration/time limits via AskUserQuestion. - Skip Logic: Skips re-asking items already explicit in the request, and starts immediately if all four points are already specified. - Scope Discipline During Execution: Requires pausing to confirm before touching out-of-scope files, and mandates attaching verification evidence (screenshots, logs) to each iteration report. - Use Case: A user says "autonomously improve this UI." The Skill triggers one confirmation round covering what "improved" means, which files may change, how each iteration is verified, and the iteration cap, then work proceeds without mid-task ambiguity. ## Quick Start When you receive a vague autonomous request like "improve this as you see fit," run the task-contract confirmation before starting any work.

Frequently Asked Questions about task-contract

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

FAQPage Schema
How do I prevent correction loops on vague AI task requests?▼

Confirm four points before starting: acceptance criteria, editable scope, verification method with reporting cadence, and iteration or time limits. Fixing these as a contract in one round prevents post-hoc correction loops on open-ended requests.

When should acceptance criteria be confirmed before autonomous work?▼

Confirm whenever the request lacks explicit done-criteria, such as "improve this" or "handle it autonomously." If the request already specifies all four contract items, skip confirmation and start immediately.

What should an AI agent do when work exceeds the agreed scope?▼

Stop and confirm with the user before touching out-of-scope files or creating unexpected new files. The contract discipline forbids silently expanding scope mid-task.

Does task-contract handle deep planning or design interrogation?▼

No. Deep plan and design questioning is delegated to the grill-me or grill-with-docs skills. This skill only fixes the execution contract: criteria, scope, verification, and limits.

What happens when the iteration limit is reached?▼

The agent stops iterating, reports the current state with verification evidence, and asks the user for a decision. It does not continue past the agreed cap on its own.