skills.rest
tech-leads-club
tech-leads-clubOFFICIAL
Explore all skills by this author →
WebsiteGitHub

not-your-babysitter

Drives autonomous end-to-end task execution with evidence-based verification and minimal user interruptions.

5.1k|468|Updated Jan 19, 2026
Search Tags:#verification#autonomous agent#evidence-based#AI agent behavior#Software Engineering#Task Automation#interruption policy
One-click install
npx skills add https://github.com/tech-leads-club/agent-skills --skill not-your-babysitter
Or copy as Structured Prompt for Agentâ–¼
Please help me install this Agent Skill.
Skill: not-your-babysitter
Source: https://github.com/tech-leads-club/agent-skills/tree/main/packages/skills-catalog/skills/%28development%29/not-your-babysitter
Command: npx skills add https://github.com/tech-leads-club/agent-skills --skill not-your-babysitter

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

AI agents often over-ask questions, guess at facts, claim work is done without proof, or spin in loops on failures, forcing users to babysit every step. This Skill turns the agent into an autonomous senior operator that verifies claims against real evidence, decides on reasonable defaults, and only interrupts for destructive actions, true dead-ends, or costly ambiguity.

Core Features & Use Cases

  • Evidence-first operation: Every claim about the outside world is verified via current web search, and every codebase claim is checked against the actual source; numbers are pulled from real sources, never memory.
  • Three-stop interruption policy: The agent stays silent and keeps working, surfacing only irreversible actions, exhausted dead-ends, or outcome-changing ambiguity, using a compact BLOCKED/TRIED/NEED format.
  • Adjustable autonomy levels: Switch between paired, solo, and heads-down modes to control how visible and talkative the agent is without loosening verification.
  • Use Case: Tell the agent to work autonomously on a multi-step engineering task, such as fixing a failing build and deploying it, and it will verify each step, keep state notes on disk, and report only the finished, proven result.

Quick Start

Tell the agent to enable not-your-babysitter mode and solve your task autonomously without hand-holding.

Frequently Asked Questions about not-your-babysitter

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

FAQPage Schema
How do I make an AI agent work autonomously without constant questions?â–¼

Enable this mode and the agent decides underspecified details using the most reasonable default, states the assumption in one line, and keeps working. It only stops for destructive actions, true dead-ends, or ambiguity that changes the outcome.

How does the agent verify its claims instead of guessing?â–¼

Claims about the outside world are checked with web search pinned to the current month and year, while codebase claims are verified by reading the actual source. Numbers are pulled from real sources in the current run, never from memory.

When will the agent interrupt me during autonomous work?â–¼

Only three events trigger an interruption: an irreversible or destructive action, a dead-end after exhausting all evidence sources, or ambiguity that is costly to get wrong. Everything else is handled silently.

What autonomy levels does this mode support?â–¼

Three levels are available: paired for close collaboration with check-ins, solo as the default with minimal interruptions, and heads-down for maximum autonomy. Verification rigor stays identical across all levels.

When should I not use autonomous agent mode?â–¼

Avoid it when you explicitly want a tutorial, a verbose walkthrough, or open-ended brainstorming. The mode is optimized for execution and short evidence-backed answers, not teaching or exploratory discussion.

Related Skills

Explore More Software Engineering Skills→
ClementATHClementATH

mcp-setup

Configure MCP servers for Claude Code with CLI commands and verification.

Community
Intermediate
kk20300113-pngkk20300113-png

ship

Automate code release workflows from branch merging to pull request creation.

Community
Advanced
jleechanorgjleechanorg

requesting-code-review

Run static scans, quality gates, and independent subagent reviews before commits.

Official
Intermediate
jayteealaojayteealao

diataxis:tutorial

Create step-by-step tutorials with goals, prerequisites, checkpoints, and recaps.

Community
Intermediate