Prompt Packs › AI Coding Agent Master Pack
AI Coding Agent Master Pack
20 structured prompts and 6 workflows for using Claude Code, Codex, Cursor and similar agents safely.
$19 one-time · no subscription
Download the free sample (3 prompts, ZIP)Purchasing isn't open yet. The free sample is available now: no email, no account.
Who it's for
Developers and technical founders who use AI coding agents on real codebases.
Problems it solves
- Agents edit before they understand the code
- Invented APIs, files and flags
- Huge unscoped diffs and drive-by refactors
- "Tests pass" without proving anything
- No clean handoff between sessions or agents
What's inside
- 20 prompts in four modes: read-only, planning, scoped implementation, validation
- 6 workflows: new feature, bug to fix, safe refactor, release day, regression hunt, taking over a repo
- Templates: agent rules snippet, SCOPE block, acceptance criteria, handoff
- 2 filled-in examples (illustrative)
- All prompts as JSON + prompt index
- PDF guide (9 pages)
Written for
Claude Code · OpenAI Codex · Cursor · Cline · Aider · Other agents that read a repo
Plain-text prompts; compatibility with each agent not yet verified by us.
The 20 prompts
| # | Prompt | Mode |
|---|---|---|
| 01 | Repository discovery & orientation | read-only |
| 02 | Read-only architecture analysis | read-only |
| 03 | Requirements clarification | planning |
| 04 | Feature implementation plan | planning |
| 05 | Safe scoped implementation | implementation |
| 06 | Minimal-diff bug fix | implementation |
| 07 | Debugging & root-cause analysis | read-only |
| 08 | Multi-file change coordination | planning |
| 09 | Refactor with explicit permission | implementation |
| 10 | Test planning | planning |
| 11 | Validate AI-generated code | validation |
| 12 | Security review of a change or area | read-only |
| 13 | Dependency assessment | read-only |
| 14 | Performance investigation | read-only |
| 15 | Deployment readiness check | validation |
| 16 | Change summary & handoff | read-only |
| 17 | Regression analysis | read-only |
| 18 | Rollback planning | planning |
| 19 | Prompt iteration for agent tasks | planning |
| 20 | Handling uncertainty & missing context | read-only |
Example prompt
From 01 · Repository discovery & orientation (read-only). Each prompt file also explains inputs, deliverables, validation and failure modes.
# Task: Repository discovery & orientation ## Mode READ-ONLY MODE. You may read files, search, and run commands that do not change state ... This prompt never authorizes edits. ## Steps 1. List top-level structure and identify language(s), frameworks, package managers and build tools from manifest files (package.json, pyproject.toml, go.mod, Cargo.toml, etc.). 2. Identify entry points (main files, server start, CLI, routes) and how the app is run locally, citing scripts/Makefile targets. 3. Identify test setup (framework, location, command) and CI config. ... ## Acceptance criteria - [ ] Every command and path cited exists in the repo - [ ] Working tree identical to the recorded initial state (no files created, changed or deleted) - [ ] Unknowns are listed, not guessed
What a good answer looks like (illustrative, not a recorded run)
Repo map: `app/` (routes), `lib/billing/` (Stripe helpers, lib/billing/stripe.ts:12) ... Run: `pnpm dev` (package.json scripts.dev) ... Risks: no tests under lib/billing ... Questions: which Stripe webhook events are enabled?
What you receive
- ZIP download: README, QUICKSTART, LICENSE, CHANGELOG, VERSION
- prompts/, workflows/, templates/, examples/, references/
- 9-page PDF guide
Formats: ZIP with .md, .txt, .json and .pdf. Works offline; no account needed.
Version: 1.1.0 · SHA-256 (pack ZIP): 056ae039246712a13c369efd31dbac8fd6b771d0a46c3c2d5b279d17cea4b468
License summary
Use and modify for your own and your organization's internal work, including commercial projects. Don't resell, redistribute or publish the pack. One purchase covers one person.
Refunds
Proposed: full refund within 14 days if it isn't useful to you. Email [email protected].
FAQ
Do I need an account or subscription?
No. You get a ZIP of plain files to keep. No logins, dashboards or usage limits.
Which agents does it work with?
The prompts are plain text written for agents that can read a repository, such as Claude Code, OpenAI Codex, Cursor, Cline or Aider. We haven't tested them with those agents yet, so we don't claim specific compatibility.
Have the prompts been tested?
Partly, and we publish the details. In v1.1.0 we ran 6 scenarios in a disposable test repo with our own AI assistant: 5 passed and 1 was a partial pass. We tightened that prompt, but haven't retested it with our assistant yet; a small local model failed the retest. Because that assistant also wrote the prompts, treat this as a consistency check, not independent proof. Small local open-source models (1.5B to 3B) failed most scenarios. We haven't yet tested with mainstream coding agents. The full results are in references/test-results.md.
Can I use it at work?
Yes, for your own and your organization's internal work, including commercial projects. You can't resell or redistribute it. For team-wide use, email us.
Do I get updates?
Updates within version 1.x are free for buyers.
What's the refund policy?
Proposed: if the pack isn't useful to you, email [email protected] within 14 days of purchase for a full refund. (Final policy will be confirmed before checkout opens.)
This is a ShipWithAI product. It doesn't affect our tool rankings.