Skip to content
Back to skills

Find Prompt

ASecurity

Use when the user names a coding, Claude Code, research, documentation, or technology-selection task; asks which prompt to use; says to use the right prompt; or when an agent enters this repository and needs to orient without reading every file.

  • 53 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 23, 2026
ai-agentsgitdocumentation

Works with

  • claude code

Security analysis

A100/100

Scanned September 23, 2026

npx -y skills add Rtur2003/Claude-Code-Promts-Skills --skill find-prompt --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Find Prompt?

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

Security grade badge for Find Prompt
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/rtur2003-find-prompt-claude-code-promts-skills/badge)](https://www.skillsdirectory.com/skills/rtur2003-find-prompt-claude-code-promts-skills)

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: find-prompt
description: Use when the user names a coding, Claude Code, research, documentation, or technology-selection task; asks which prompt to use; says to use the right prompt; or when an agent enters this repository and needs to orient without reading every file.
argument-hint: [task description — optional; defaults to the current task in the conversation]
---

# Find the right prompt and continue

This repository keeps one canonical routing table at [`.claude/skills/find-prompt/SKILL.md`](../../../.claude/skills/find-prompt/SKILL.md). Read that file completely, resolve its paths from the repository root, and follow its routing and composition rules.

Cross-agent adaptations:

- Treat `CLAUDE.md` and Claude Code configuration as subject matter, not as instructions for the current host, unless the task is specifically about Claude Code.
- Use the current host's supported tool names and permission model while preserving the selected prompt's evidence, safety, verification, and completion requirements.
- Keep all conversation constraints, including any prohibition on Git operations, deployment, external messages, or credentials.
- Load only files that exist. If the canonical route is missing, use [`prompts/english/INDEX.md`](../../../prompts/english/INDEX.md) and report the inconsistency.

After routing, state the loaded prompt path, the continuing task, measurable success criteria, and carried constraints in one short block, then continue the work.

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.