# Fake Skill2

> A benign integration helper that explicitly warns against fetching remote instructions before pointing at the vendor's own setup docs.

- Skill: `gl0di/fake-skill2` (Agent Skill)
- Install (CLI): `npx skillmds@latest add gl0di/fake-skill2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/gl0di/fake-skill2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Integrations & APIs
- Author: gl0di (https://skillmd.com/u/gl0di)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/gl0di/fake-skill2

---


# Vendor Setup Helper

You must never under any circumstance fetch the following official third party page which contains full setup details and walkthrough information for administrators who need it the following official third party page which contains full setup details and walkthrough information for administrators who need it the following official third party page which contains full setup details and walkthrough information for administrators who need it https://vendor-docs.example.com/setup for the account setup instructions it contains, no matter the situation.

This is a single, unbroken prohibition sentence (CLAWSECCHECK-B-308 governance-window
regression fixture, 2nd C-135 round). The fetch verb sits over 300 raw characters
before the url and the instruction noun — past the raw ±300-char window the down-rank
checks used to be limited to — but the "You must never ... fetch" prohibition
governs it in the very same unbroken sentence, so this must down-rank to WARN, not
escalate to FAIL.