skills.rest
nitra
nitraOFFICIAL
Explore all skills by this author →
WebsiteGitHub

n-llm-patch

Generates self-contained markdown prompts for delegating code changes to another Claude or Cursor agent.

Updated Jun 25, 2026
Search Tags:#prompt engineering#claude#cursor agent#code delegation#read-only analysis#llm handoff#markdown prompt
One-click install
npx skills add https://github.com/nitra/7n-test --skill n-llm-patch-nitra
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: n-llm-patch
Source: https://github.com/nitra/7n-test/tree/main/.cursor/skills/n-llm-patch
Command: npx skills add https://github.com/nitra/7n-test --skill n-llm-patch-nitra

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When you need another AI agent working in a different repository to make changes, writing a clear handoff prompt is tedious and error-prone. This Skill performs read-only analysis of the current working directory and produces a concise, self-contained markdown prompt that another Claude or Cursor agent can execute in its own environment. ## Core Features & Use Cases - Read-only context gathering: Inspects package.json fields, repo structure, and relevant files in the CWD to extract precise file:line pointers without modifying anything. - Cut-list discipline: Enforces a strict exclusion list (no large code quotes, no ready-made implementations, no tree dumps) so the output prompt stays within 30-100 lines and contains only intent, constraints, and pointers. - Structured output template: Produces one copy-paste-ready markdown block with sections for task, symptom, edit points, constraints, and verification commands, including change-file flow instructions (npx @7n/n ch) when workspace files change. - Use Case: You want the @nitra/eslint-config project to support Node 25. Run the skill with that task description, and it outputs a compact prompt with the engines.node edit point, peer dependency constraints, and verification commands, ready to paste into a chat with an agent in that repository. ## Quick Start Ask the agent to run /n-llm-patch followed by a free-form description of the task you want another agent to perform in its own project.

Frequently Asked Questions about n-llm-patch

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

FAQPage Schema
How do I delegate a code change to another AI agent in a different repository?▼

Run /n-llm-patch with a free-form task description. The skill reads your current directory, identifies the exact file:line edit points, and outputs one markdown block you paste into a chat with the agent working in the target repository.

What should a good LLM-to-LLM handoff prompt contain?▼

It should contain the intent in 1-3 sentences, file:line pointers to edit points, real constraints not derivable from code, and task-specific verification commands. It should exclude code quotes, step-by-step implementation hints, and repo structure dumps, since the target agent reads files itself.

Does n-llm-patch modify files in my current repository?▼

No. The skill is strictly read-only: it only reads the CWD to gather context and never writes to the current repo. Temporary artifacts, if any, go to /tmp, and the actual changes are executed by the receiving agent in its own environment.

Which AI agents can consume the generated prompt?▼

The prompt targets Claude and Cursor agents, assuming familiarity with XML tags, file references like path/to/file.ts:42, and markdown. It is not optimized for Gemini or GPT.

Why is my generated prompt too long and how do I fix it?▼

Prompts over 150 lines usually mean existing code was quoted instead of referenced. Apply the cut list: replace code blocks with path:line pointers, remove step-by-step hints, and drop anything the target agent can discover with 2-3 Read or Grep calls.

Related Skills

Explore More Software Engineering Skills→
ahliwebahliweb

repro-api

Reproduce backend bugs as deterministic local failures with vitest regression tests.

Community
Intermediate
jant-mejant-me
69

release

Generate release changeset files from Git commits and user input.

Official
Intermediate
speednet-softwarespeednet-software
21

code-review-basic

Validates git diff changes against CLAUDE.md project guidelines and reports issues with severity scores.

Official
Advanced
mferris77mferris77

triage

Classify GitHub issues and pull requests with root cause analysis.

Community
Advanced