Skillget

remy-skill-recipes

agent
Guvenlik Denetimi
Gecti
Health Gecti
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 10 GitHub stars
Code Gecti
  • Code scan — Scanned 2 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested
Purpose
This project provides a collection of structured prompt workflows and recipes designed for software engineering LLM agents. It follows the SKILL.md open standard to help with tasks like debugging, documentation, code review, and automation.

Security Assessment
Overall risk: Low. The tool is fundamentally a collection of text-based prompt templates rather than executable software. The automated code scan reviewed 2 files and found no dangerous patterns, hardcoded secrets, or requests for elevated system permissions. Because these are prompt recipes, they do not inherently access sensitive data, execute shell commands, or make external network requests. However, because the quick start guide suggests installing via `npx`, users should ensure they understand what the npm package executes on their local machine before running the command.

Quality Assessment
The project demonstrates solid quality and active maintenance. It was updated very recently (0 days ago) and operates under the permissive and standard MIT license. While community trust is currently modest at 10 GitHub stars, this is typical for niche developer utilities. The repository is well-documented, clearly outlining how to use the skills across various popular AI coding environments like Cursor, Claude Code, and VS Code.

Verdict
Safe to use.
SUMMARY

Engineering-grade skill recipes for LLM agents. 7 structured prompt workflows following the SKILL.md open standard.

README.md

Remy Skill Recipes

Engineering-grade skill recipes for working with LLMs.

This repository is a curated collection of reusable, structured prompt workflows
designed for software engineering tasks — debugging, documentation, review,
research, automation, and system analysis.

Each skill follows the SKILL.md open standard —
compatible with Cursor, Claude Code, VS Code + Copilot, Codex, Gemini CLI, and other SKILL.md-compatible agents.

Quick Start

Browse

Skill Type Description
architecture-spec Execution Risk-based architecture doc generation with auto-leveling (A/B/C)
change-reaudit Execution Re-audit code changes for side effects, regressions, and edge cases
competitive-feature-benchmark Execution Compare competing products at the UX and interaction level
docs-finalize-and-commit Execution Finalize documentation with convention discovery and consistency checks
finalize-and-commit Execution Finalize code for production — dedup, hardcode audit, clean commits
notion-format Execution Auto-detect content type and format rich Notion documents
oss-code-analysis Execution Analyze OSS repos at the code level (compare or deep-dive mode)
ux-sentinel System Auto-detect recurring UX concepts and persist to Notion DB

Install a skill

Install all skills at once from the skills.sh marketplace:

npx skills add comsky/remy-skill-recipes

Or copy a single skill manually:

# Cursor
cp -r skills/change-reaudit ~/.cursor/skills/

# Claude Code
cp -r skills/change-reaudit ~/.claude/skills/

# VS Code + Copilot (auto-discovers .claude/skills/ or use dedicated path)
cp -r skills/change-reaudit ~/.copilot/skills/

The agent will automatically discover and activate the skill when a matching task appears.

Use directly

  1. Identify whether your task needs:
    • An Execution Skill (single-run workflow), or
    • A System Skill (persistent / automation behavior).
  2. Read the Inputs Required section carefully.
  3. Provide complete context.
  4. Validate output using the skill's checklist.

Most bad outputs come from incomplete inputs.

Skill Format

Every skill is a folder containing a SKILL.md file with YAML frontmatter:

---
name: change-reaudit
description: >
  Re-audit code changes to identify side effects, regression risks,
  and unhandled edge cases before merging or deploying.
license: MIT
compatibility:
  - Claude Code
  - Cursor
metadata:
  type: execution
  category: review
  maturity: stable
  estimated_time: 10 min
---

Agents read only name and description during discovery (~100 tokens).
The full markdown body loads on activation (<5000 tokens).

Skill Types

1. Execution Skills

Single-run structured workflows.

Used for:

  • Code review
  • Change auditing
  • Competitive benchmarking
  • Documentation writing
  • Refactoring validation

These skills:

  • Require explicit inputs
  • Produce structured outputs
  • Contain guardrails and failure patterns
  • Include realistic examples

2. System Skills

Persistent or automation-oriented behaviors.

Used for:

  • Continuous concept detection
  • Knowledge tracking
  • External DB synchronization
  • Conversation-wide logic

These skills:

  • Define activation rules
  • Maintain state (conversation or external DB)
  • Specify side effects explicitly
  • Include operational guardrails

Skill Categories

Category Description
review Change audits, regression analysis
research Competitive feature analysis
cleanup Refactor and commit structuring
documentation Architecture specs, design docs
automation Persistent or DB-connected skills

Skill Structure

Execution Skill Structure

  • Purpose
  • When to Use
  • When NOT to Use
  • Inputs Required
  • Output Format
  • Procedure
  • Guardrails
  • Failure Patterns
  • Examples (minimum 2)

System Skill Structure

  • Purpose
  • Scope (triggers / non-triggers)
  • Inputs / Signals
  • Core Behavior (Detection → Decision → Action)
  • Output / Side Effects
  • Guardrails
  • Failure Patterns
  • Examples

Skill Maturity Levels

  • Draft — experimental
  • Stable — reliable for repeated use
  • Production — validated in real workflows

Model Assumptions

Skills are model-agnostic.

Assume:

  • A reasoning-capable LLM
  • Structured context input
  • Long-context support preferred

Philosophy

LLMs are unreliable without structure.

Structure reduces:

  • Hallucination
  • Ambiguity
  • Context loss
  • Overconfidence errors

Prompting is engineering.

System skills extend this philosophy
into persistent decision-memory patterns.

Contribution

This repository is primarily maintained for personal reuse,
but high-quality pull requests are welcome.

Rules:

  • Use the correct template (_template/execution-template.md or _template/system-template.md)
  • Follow the SKILL.md standard (YAML frontmatter required)
  • Include realistic examples
  • Document guardrails and failure patterns

License

MIT

Yorumlar (0)

Sonuc bulunamadi