home/roleplay/starting-prompt

Starting Prompt

GPTClaudeGemini··417 copies·updated 2026-07-14
starting-prompt.prompt
# Starting Prompt - Paste-and-Go

The whole point: clone the repo, open Claude Code in the folder, type one command,
answer one question the very first time, and the framework does the rest. It runs a
batch of evals and sends each result the moment it finishes.

How results come back depends on that one first-run answer:

- **Give a Telegram bot token + chat id -> results arrive in Telegram** as document
  attachments, one per eval. No git push, no GitHub access needed at all. This is the
  easiest path - nothing to copy back.
- **Press ENTER to skip -> results push to a git branch** and you copy the printed
  branch name back to Damian (needs GitHub push access).

If you do not have Claude Code, see [METHOD.md section 12](../METHOD.md#12-for-users-without-claude-code-agents-lite-path)
for the lite path.

---

## Prerequisites

1. **Claude Code installed** (the agent harness that runs the orchestrator and its
   sub-agents).
2. **git installed** (used to clone the repo either way).
3. **GitHub push access - ONLY if you skip Telegram.** If you have a Telegram bot
   token + chat id, you do NOT need any GitHub auth - results are delivered straight
   to Telegram and nothing is pushed. Only the git-push (no-token) mode needs push
   access (via `gh auth login` or a clone token). You can never touch `master`
   either way (it is protected, by design).

---

## The four actions

### 1. Clone the repo

when to use it

Community prompt sourced from the open-source GitHub repo DamianBuilds-ai/model-gauntlet (MIT). A "Starting Prompt" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.

tags

roleplaycommunitygeneral

source

DamianBuilds-ai/model-gauntlet · MIT