Communitygithub.com

Bum-Boo/hermes-skill-library

Curated Hermes Agent skill library by bumboo

hermes-skill-library 是什麼?

hermes-skill-library is a Claude Code agent skill that curated Hermes Agent skill library by bumboo.

相容平台~Claude Code~Codex CLI~Cursor
npx skills add Bum-Boo/hermes-skill-library

在你喜歡的 AI 中提問

開啟一個已預先載入此 Agent Skill 的新對話。

說明文件

gstack-safe

Overview

Use this as the umbrella router for a conservative, evidence-first workflow family. It helps an agent decide whether to write a spec, review a proposed change, or investigate a failure before editing.

When to Use

Use when a request needs one of these modes:

  • Spec: the goal is unclear, broad, or not yet ready for implementation.
  • Review: the user asks whether a plan, prompt, diff, or design is safe and complete.
  • Investigation: something failed and the root cause is not known.

Non-Triggers

Do not use this skill to bypass normal safety requirements. It does not grant permission to:

  • deploy, publish, push, merge, or delete;
  • access private accounts, cookies, production data, or credentials;
  • run browser actions against a user account;
  • claim success without verifying current evidence.

Routing

  1. If the user wants a bounded implementation plan, load gstack-spec.
  2. If the user wants critique or risk analysis, load gstack-review.
  3. If the user provides errors, logs, failing tests, or unclear behavior, load gstack-investigate.
  4. If the request includes external side effects, stop and ask for explicit approval or route to a more specific skill.

Default Guardrails

  • Inspect current evidence before proposing changes.
  • Separate facts, assumptions, and recommendations.
  • Prefer small reversible steps.
  • Keep destructive or externally visible actions behind explicit approval.
  • Report what was verified and what remains unknown.

Output Contract

When used as a router, respond with:

  • selected workflow;
  • reason for the selection;
  • required inputs, if any;
  • immediate next safe step.

相關技能