Code & Agents
For developers who work with Claude Code, Codex and Cursor: root-cause debugging, merge verdicts on diffs, tests first, refactoring plans and one AGENTS.md for every agent.
33 prompts and 85 agent skills · 10 free to read in full
Tested CLAUDE.md / AGENTS.md rules and prompts from 4 apps and 700k lines I shipped with AI. Real numbers, free starter prompts.
One-time payment · VAT included · files you keep · no account
Codex, GPT-6.1 Sol, Ultra. The same task in both runs: a 15-second Blender video at 720p.
| Setup | Tokens used | 5-hour limit used |
|---|---|---|
| Stock Codex with 3 subagents | 3,918,231 | +14 pp |
| My AGENTS.md, no subagents | 1,874,977 | +8 pp |
52% fewer tokens · 43% smaller increase in the 5-hour limit
One test pair. Run took 2 min longer.
New drops every week.
Read them in full, or download all 10 free items as one zip. No account needed.
The example from Grok Trend Angler, one of the free prompts, as it appears on its page: what you give it and the answer format it asks the model for. It is an illustration written by Meevra; to see what it does with your own material, copy the prompt and run it on your model.
Input: topic: four-day work week pilots · product: Notewell, a meeting-notes app that turns calls into action lists; goal: trial sign-ups from team leads · knowledge: our 8-person team cut recurring meetings from 9 to 4 a week last quarter · posts: 12 pasted posts from founders and managers
POST FIRST: Post now, as a reply to the manager in your pasted posts who doubts a team can cover a four-day week. Your own meeting numbers answer that doubt, and team leads are the people reading that thread.
Four-day weeks don't fail because of the day off. They fail because the meetings stay.
Our 8-person team cut recurring meetings from 9 a week to 4 last quarter. Cut meetings first, then the day. I build Notewell, which turns calls into action lists. (249 characters)
If challenged: your own meeting count from last quarter.
TREND CHECK: Active, based on the posts you pasted: two pilot results, the manager's doubts about coverage, a founder's "never again" post.
ALSO GOOD
Useful number: One weekly 60-minute meeting with 8 people is 8 person-hours. On a 32-hour week, that is a quarter of one person's week. Every week. (132 characters)
If challenged: arithmetic from the inputs shown (60 min × 8 = 8 hours; 8 ÷ 32 = 25%); no survey figure claimed.
Later today (optional): reply with the first recurring meeting your team cut and what took its place.
Illustrative example written by Meevra to show the target format; your model's wording will differ.
Each pack is 12 €, VAT included. Each pack also holds the free items of its job.
For developers who work with Claude Code, Codex and Cursor: root-cause debugging, merge verdicts on diffs, tests first, refactoring plans and one AGENTS.md for every agent.
For building apps in Lovable and Bolt: a spec the builder can follow, tables and access rules planned first, changes in a safe order and a pre-launch audit.
For founders and freelancers who run the business side themselves: weekly reviews, meeting decisions, contracts, SOPs, hiring, pricing and email.
For content that has to be found and read: SEO briefs, Search Console fixes, landing-page copy, brand voice, localization and posts for X.
For decisions you have to back up: answers from your data with the numbers shown, errors found in financial models, documents compared and papers turned into findings you can check.
For AI video in Veo, Kling, Runway and Seedance: shot lists planned to cut together, one frozen character description across clips, dialogue cut to fit each clip and 9:16 ads.
For product and ad images in Midjourney, FLUX, Nano Banana and GPT-Image: one style block for a whole series, listing sets checked against your photos, text cut to fit and ad variants.
What an item does not do is make the model smarter. For many jobs a plain request works well; what you pay for is the ready checklist, format and workflow, and the time you save not writing them yourself.
One-time purchases. No account and no subscription.
No account or email needed
VAT included · one-time
VAT included · one-time
All-Access 39 € vs 7 packs × 12 € = 84 € · less than buying 4 or more packs one by one
Checkout by card (Stripe) or PayPal, or via Payhip. These are digital files: at checkout you agree that the download starts right away, so the 14-day right of withdrawal ends once you can download. Faulty files are still covered by the legal guarantee, and if you get stuck setting something up, we help. Questions before you buy
Stuck? Email support@meevra.com — we'll help you set it up.
I'm Dainius, a solo builder from Lithuania. I've shipped 4 apps (~700k lines) with Claude & Codex and publish what actually saves quota.
You can, and for many jobs a plain request works well: the model does the work either way, and a Meevra item does not make it smarter. What you buy is the setup you would otherwise write and remember yourself: the job's checklist, the questions to settle first, a consistent answer format you can reuse, and notes on when to use it, when to skip it and how to run it on your model. A skill goes one step further: you install its SKILL.md once, and your agent reuses it whenever the job comes up.
For example, instead of typing "Help me review my week" and adding details as you remember them, you paste the free Weekly Business Review prompt with your notes. It compares the week with last week's file, asks for a decision on each time drain (keep, limit, fix the cause, hand off or stop) and ends with one number to check next Friday, in the same format every week. You can read it in full and run it on your own notes before you buy anything.
If you already keep your own checklists and prompts for these jobs, an item saves you less. Each item page says when to skip it.
Yes. Each job pack is 12 €, VAT included, one-time, and holds every prompt and skill for one job, plus the free items of that job. If you need 4 or more of the 7 jobs, All-Access (39 €) costs less than buying the packs one by one.
Each item page lists the models and tools it was written for and says how to run it there: where to paste it (a chat, a Project, a Gem, a skill folder or an agent rules file) and which effort or thinking level to use. Those model facts come from our model pages, such as Claude and ChatGPT, with a source and a check date for each.
Prompts are plain text, so they also work in other chat tools. Skills use the open Agent Skills format, and 80 of the 85 skills also work pasted into a normal chat; their page says so. The surest check is your own: run a free item on your model before you buy.
Yes, if you can copy and paste. A prompt is text you paste into a chat, after replacing the [PLACEHOLDERS] in capitals with your own details. A skill is a folder you add where your AI tool loads skills, or you paste its SKILL.md into a chat. Every item says how to run it, and every pack comes with a QUICKSTART.
Start with the free items: they show their full text, so you can see exactly what you would be working with. Stuck? Email support@meevra.com and we'll help you set it up.
These are digital files. At checkout you agree that the download starts right away and confirm that you then lose the 14-day right of withdrawal, so once the files are available there is no refund for a change of mind. If you were not asked for that consent, you keep the right to withdraw for 14 days.
If a file is faulty, incomplete or not as described, the 2-year legal guarantee still applies: we fix it, or you get a price reduction or a refund. And if an item does not work the way you expected, write to us first: most problems are a setup step, and we help you with it. The details are in sections 5 and 6 of the terms. Before you buy, you can read every free item in full and see each paid item's use cases and example.
Search the library or open a pack. Each item page says when to use it and when to skip it, with an example.
Prompts are text you paste into a chat or a project. Skills are folders with a SKILL.md file in the open Agent Skills format.
Each item says where to paste it and which effort or thinking level to use. Stuck? Email us and we help you set it up.
Model facts checked 15 Oct 2026.