output-skill

Enforces complete, unabridged LLM output by banning placeholder patterns and truncation.

Updated Mar 28, 2026
One-click install
npx skills add https://github.com/po4yka/blog --skill output-skill-po4yka
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: output-skill
Source: https://github.com/po4yka/blog/tree/main/.claude/skills/output-skill
Command: npx skills add https://github.com/po4yka/blog --skill output-skill-po4yka

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? LLMs often truncate long responses, insert placeholder comments like "// rest of code", or skip deliverables to save tokens. This Skill overrides that default behavior so every requested file, function, or section is generated in full. ## Core Features & Use Cases - Placeholder Ban: Hard-fails on patterns like "// ...", "// TODO", "for brevity", and "the rest follows the same pattern" in both code and prose. - Deliverable Counting: Scopes the request, locks the expected deliverable count, and cross-checks output against it before responding. - Token-Limit Handling: When output approaches the token limit, it stops at a clean breakpoint and emits a resumable pause marker instead of compressing or skipping content. - Use Case: Ask for a full multi-file implementation (e.g., 5 React components with tests) and receive every file completely, with no skeletons or omitted sections. ## Quick Start Apply the output-skill rules and generate the complete implementation for all requested files without any placeholders or omissions.

Frequently Asked Questions about output-skill

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

FAQPage Schema
How do I stop an LLM from truncating code output?

Apply explicit output rules that ban placeholder patterns like "// rest of code" and require every requested deliverable to be generated in full. This Skill enforces that behavior with a banned-pattern list and a deliverable cross-check before responding.

How to get complete multi-file code generation from an AI?

Scope the request by counting the expected files or functions, then require the model to verify that count before finalizing. If the token limit is reached, output pauses at a clean breakpoint with a resume marker instead of compressing remaining files.

What happens when the response hits the token limit?

The output stops at a clean breakpoint such as the end of a function or file, then emits a pause marker stating progress and the next section. Sending "continue" resumes exactly from that point with no recap or repetition.

Which placeholder patterns are banned in generated code?

Banned patterns include "// ...", "// rest of code", "// TODO", "// similar to above", and prose shortcuts like "for brevity" or "the rest follows the same pattern". Any of these in output is treated as a hard failure.

When should I not use full-output enforcement?

Avoid it for exploratory brainstorming, quick prototypes, or when you only need a sketch of an approach. The enforcement optimizes for completeness, so it produces longer responses than necessary for lightweight tasks.