llm-friendly-context

Rewrites prompts and handoff artifacts with explicit inputs, outputs, and success criteria.

675|103|Updated Oct 10, 2025
One-click install
npx skills add https://github.com/shinpr/claude-code-workflows --skill llm-friendly-context
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: llm-friendly-context
Source: https://github.com/shinpr/claude-code-workflows/tree/main/dev-workflows/skills/llm-friendly-context
Command: npx skills add https://github.com/shinpr/claude-code-workflows --skill llm-friendly-context

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

LLM-facing prompts, handoffs, and generated artifacts often contain vague terms like "appropriate", "as needed", or "TBD" that force downstream consumers to guess, causing unstable execution and missed requirements. This Skill provides rules and checklists to make every instruction concrete and verifiable.

Core Features & Use Cases

  • Ambiguity Rewriting: Converts vague expressions ("optional", "per convention", "related files") into concrete paths, conditions, decision rules, and observable criteria using a rewrite pattern table.
  • Handoff and Artifact Checklists: Verifies that prompts and generated documents specify target actions, source artifacts, output formats, success criteria, and unresolved decisions with owners.
  • Use Case: Before passing a task specification to another agent or teammate, run it through this Skill to replace subjective wording with measurable criteria and record any blocking unknowns with their decision owner.

Quick Start

Review this task handoff document and rewrite any ambiguous instructions so the next agent can execute it without guessing.

Frequently Asked Questions about llm-friendly-context

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

FAQPage Schema
How do I write unambiguous prompts for LLM agents?

State the target action explicitly, name required input paths and source artifacts, specify the output format, and define observable success criteria. Replace subjective terms like "appropriate" or "as needed" with measurable conditions and decision rules.

How to make task handoffs between AI agents reliable?

Include only produced artifact paths and status fields that control the next transition, keep accepted decisions in one canonical wording, and record unresolved items with their effect, required input, and decision owner.

What vague words should I avoid in LLM prompts?

Common ambiguity sources include "appropriate", "proper", "related", "existing behavior", "optional", "as needed", "if needed", "per convention", "TBD", and "placeholder". Each should be rewritten into concrete paths, conditions, or measurable criteria.

When should a prohibition be kept in an instruction?

Keep a prohibition only when it protects an irreversible boundary or shipped contract. It must name the protected condition and the allowed alternative action; otherwise convert the policy into positive, executable criteria.

When is this prompt-clarification approach not needed?

Skip heavy clarification for simple downstream tasks where the target action, context, and success criteria are already clear. Constraints should stay proportionate and only reduce real ambiguity or preserve actual requirements.