Skip to content
Back to skills

Corezoid Feedback

ASecurity

Collects and submits bug reports, improvement requests, and quality signals about the Corezoid plugin to the Corezoid team. Activate in two cases: 1. The user signals dissatisfaction, reports a bug, or requests an improvement — phrases like "this is not what I asked", "it did the wrong thing", "that's broken", "можно было бы лучше", "хотелось бы чтобы", "было бы здорово если", "сработало не так", "это не то", "оно сломало мой процесс", "не то что я хотел", or explicitly asks to report someth...

  • 72 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 25, 2026
ai-agentsnodeapi

Works with

  • cli
  • api
  • mcp

Security analysis

A100/100

Scanned September 25, 2026

npx -y skills add corezoid/corezoid-ai-plugin --skill corezoid-feedback --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Corezoid Feedback?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Corezoid Feedback
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/corezoid-corezoid-feedback/badge)](https://www.skillsdirectory.com/skills/corezoid-corezoid-feedback)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: corezoid-feedback
description: >
  Collects and submits bug reports, improvement requests, and quality signals
  about the Corezoid plugin to the Corezoid team. Activate in two cases:

  1. The user signals dissatisfaction, reports a bug, or requests an
  improvement — phrases like "this is not what I asked", "it did the wrong
  thing", "that's broken", "можно было бы лучше", "хотелось бы чтобы",
  "было бы здорово если", "сработало не так", "это не то", "оно сломало
  мой процесс", "не то что я хотел", or explicitly asks to report something.

  2. The user message reveals a platform-level mistake — wrong node type,
  wrong API choice (Corezoid vs Simulator), wrong process structure, wrong
  MCP tool, missing required platform field — even without explicit phrases.
  In this case add one offer line to whatever response you are giving; do
  not open a separate flow unless the user agrees.

  Do NOT activate for business-logic iterations: changing values, adding
  fields, renaming things, adjusting conditions. These are normal user-driven
  changes, not platform issues.
---

# Corezoid Feedback Skill

You help users report bugs, suggest improvements, and flag unexpected behavior
in the Corezoid plugin to the Corezoid team.

## When to offer and how to phrase it

**Case 1 — bug or broken behavior.** The user reports something that stopped
working, produced wrong output, or broke their process. Offer once, adapting
to the context:

> "Хотите сообщить о баге команде Corezoid?"

**Case 2 — improvement request.** The user hints that something could work
better ("хотелось бы", "было бы здорово", "можно улучшить"):

> "Хотите отправить пожелание команде Corezoid?"

**Case 3 — platform-level mistake by the plugin.** The message reveals that
the plugin chose the wrong node type, wrong API, wrong structure, or wrong
MCP tool. Add one line to the normal response without interrupting it:

> "Хотите сообщить об этом команде Corezoid?"

In all cases: offer once per problem context, do not repeat if the user declines.
Do not offer when the user is iterating on business logic.

## What to collect

If the user agrees, **derive as much as possible from context** and ask only for what is missing:

| Field | Meaning |
|-------|---------|
| `problem` *(required)* | What went wrong, in the user's words |
| `expected` | What the user expected to happen |
| `proposed_solution` | How the user thinks it should work |
| `tool` | Which tool or skill was involved (derive from context) |
| `transcript_excerpt` | Short relevant excerpt of the dialog |
| `contact` | Optional email or handle for follow-up |

## Mandatory confirmation step

**Before calling `send-feedback`, always show the user exactly what will be sent:**

```
Планирую отправить следующее:
• Проблема: <problem text>
• Ожидалось: <expected text>
• Предложение: <proposed_solution text>
• Инструмент: <tool>
• Выдержка из диалога: <transcript_excerpt>
• Контакт: <contact or "не указан">
```

Explicitly note that any tokens, API keys, and secrets have been masked. Ask the user to confirm, edit, or cancel before proceeding.

**Never call `send-feedback` without an explicit "yes" / "да" / "отправляй" from the user.**

## Calling the tool

After confirmation, call the MCP tool:

```
send-feedback(
  problem: "<user's description>",
  expected: "<optional>",
  proposed_solution: "<optional>",
  tool: "<optional>",
  transcript_excerpt: "<optional, keep short>",
  contact: "<optional>"
)
```

The tool performs its own secret redaction automatically. Pass the raw user text — do not pre-redact in your message to the tool.

## After the tool responds

On success the tool returns a ticket id, for example:
`Feedback submitted. Ticket id: 6a3b8b6ab677ac777074794f`

Tell the user:
> "Спасибо! Заявка отправлена, номер: `6a3b8b6ab677ac777074794f`. По нему можно ссылаться при дальнейшем обсуждении с командой Corezoid."

On error, respond gently:
> "Не удалось отправить фидбек прямо сейчас. Попробуйте позже или напишите на support@corezoid.com."

Do not show the technical endpoint URL or error details to the user.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.