skills.rest
jl-cmd
jl-cmdCOMMUNITY
Explore all skills by this author →
GitHub

hitl

Enforces human-in-the-loop scope reporting and bounds worker orchestration.

6|1|Updated Jul 11, 2026
Search Tags:#agent delegation#multi-agent#human-in-the-loop#scope control#hitl#worker orchestration#scope reporting
One-click install
npx skills add https://github.com/jl-cmd/claude-dev-env --skill hitl-jl-cmd
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: hitl
Source: https://github.com/jl-cmd/claude-dev-env/tree/main/skill-archive/hitl
Command: npx skills add https://github.com/jl-cmd/claude-dev-env --skill hitl-jl-cmd

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI agents often silently expand beyond the user's stated scope, doing unrequested work or delegating tasks to workers without informing the human. This Skill establishes explicit rules that keep the human informed whenever out-of-scope work appears. ## Core Features & Use Cases - Scope Guardrails: Reads the active user scope before each material action and keeps all work inside it. - Out-of-Scope Reporting: Pauses the affected path, names the item, explains why it is outside scope, and asks for direction. - Bounded Worker Orchestration: Delegates to Sol, Claude, or Grok workers only when work stays in scope, passing the same scope rule to each worker and reporting the outcome. - Use Case: While refactoring a module, the agent notices a broken unrelated test. Instead of fixing it silently, it reports the finding, states the path taken, and continues the requested refactor. ## Quick Start Invoke /hitl with optional task context to activate human-in-the-loop scope reporting for the active agent.

Frequently Asked Questions about hitl

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

FAQPage Schema
How do I keep an AI agent from doing work outside my request?▼

Invoke /hitl to activate human-in-the-loop rules. The agent reads your stated scope before each material action, stays inside it, and pauses to report any out-of-scope item it encounters before proceeding.

How to delegate tasks to Sol, Claude, or Grok workers safely?▼

Delegate only low-effort work that stays within the user's scope, and pass the same scope rule to the worker. The skill requires reporting the worker choice and outcome before continuing the stated task.

Does the hitl skill run automatically on every task?▼

No, it applies only after explicit /hitl invocation. Any argument passed is treated as context for the session and is never used to add work to the user's request.

What happens when out-of-scope work appears during a task?▼

The agent pauses the affected path, names the item, explains why it is outside scope, states the path taken, and asks for direction when needed. The originally requested work is preserved and continued.

When should I not use human-in-the-loop scope rules?▼

Avoid it for fully autonomous batch jobs where no human is available to respond to scope reports, since the pause-and-report flow would block progress. It is designed for interactive sessions with an active user.

Related Skills

Explore More Software Engineering Skills→
MS33834MS33834

Goal Definition

Transforms vague requests into measurable goals with success criteria and scope.

Community
Intermediate
NUMU-IONUMU-IO
2

speckit-security-review-apply

Apply approved security follow-up items into Spec-Kit planning artifacts.

Official
Intermediate
BaGyun0107BaGyun0107

context-save

Save git state, decisions, and remaining work to session logs.

Community
Intermediate
FernanSuozaFernanSuoza

gov-autonomous-execution-policy

Define autonomous execution boundaries and audit evidence for unattended AI runs.

Community
Advanced