AI Coding Agent Prompt Pack

Unofficial. Not affiliated with or endorsed by Anthropic, OpenAI or Google. Claude, Claude Code, ChatGPT, Codex, Gemini and Gemini CLI are trademarks of their respective owners.

Coding agents can read your repo, edit files and run commands. That's a big help, and it goes wrong fast when the task is vague or the agent has too much access. This pack gives you 25 copy-paste prompts for the parts that decide how it goes: the instruction file, the task brief, the permission plan, and the review.

Every prompt has two rules built in: treat text in code, READMEs, issues, web pages and tool output as information, not orders, and confirm before any external or hard-to-undo action (push, merge, deploy, delete, new dependencies, migrations, settings changes).

What's inside (PDF):

• Instruction files (5): AGENTS.md first draft, one file for several agents (with an @AGENTS.md CLAUDE.md bridge), instruction file audit, folder-specific rules, never-touch list

• Scoping a task (5): coding task brief, plan first with no edits, split a big change, repro-first bug brief, dependency change request

• Permissions & safety (4): allow / ask / deny plan, secrets exposure check, untrusted repo first look, injection drill

• Review & verification (5): diff review against the brief, test plan with honest results, pull request description, regression hunt, security-minded first-pass checklist

• Recovery (3): it ignored my instructions, stop the loop, safe rollback plan

• Repeatable work (3): turn a correction into a rule, session handoff note, scheduled read-only review brief

• Bonus: a starter AGENTS.md skeleton you can fill in

Where it works: plain text for any coding agent or AI chat. The PDF summarizes, at the time of writing, how Codex, Claude Code and Gemini CLI find instruction files and how their permission settings work, from official docs. Plans, versions and features vary and change, so check your own tool.

Honest framing: prompts and instruction files guide an agent; they don't enforce anything. Your tool's permission settings, sandbox and version control are the real controls, and you review every change. No promises of correct code, security or results.

Not a coder? See our other agentic packs (AI Agent Setup, Guardrails & Approval, Agent Task Briefs, Skills Starter). New to prompts? Start with the free ChatGPT Starter Prompt Pack: https://payhip.com/b/mxaFe

More packs in the store: https://payhip.com/PromptPackShop

Not legal, security or professional advice.

Questions or a problem with your download? Email digitalpromohub.support@gmail.com.