Prompt packs for builders, game devs and solo founders, page 4
Tested prompt sets with [placeholders] you fill in, rules that keep the model honest, and a clear stop point before anything irreversible. Two prompts from every pack are free to read on the page.
Codex: five config rules and a five-check script for .codex/config.toml (free sample)
From the Codex bundle: five rules for .codex/config.toml (keep the sandbox on, approval on-request, network off by default, pin the model, pin servers and keep keys out), and a read-only script that checks one config file for them.
Reach Inventory: every connector your agents hold, with a job, an owner and a way out
One row for every connector, MCP server, OAuth grant, API key and automation bridge your agents can use, each tied to a job and marked necessary, convenient or leftover. Scope granted is set against the scope the job needs, keys are named by where they live and never pasted, bridges list every app behind them, and leftover access is removed with a date and proof. Agents list and propose; removals, narrower scopes, rotations and new grants wait for your yes at your own terminal, tied to the sha256 of the inventory. Fifteen read-only checks run on the inventory.
Reach Inventory: five rules and a five-check script (free sample)
From the Reach Inventory: five rules for the connectors your agents hold (a human owner, a job for every connector, leftover means removed, write access has an approver, unused access gets reviewed), and a read-only script that checks one reach list for them.
Problem Mining Loop: real complaints, graded and counted, before you build anything
Find out what one kind of buyer complains about, in their own words. Every complaint is a row with the link, the date posted, the exact quote and a salted key in place of the poster's handle, graded gripe, pain, budget or buyer. Clusters count independent posters, not posts, and leave out prompted answers and your own accounts. A cluster is proposed only when enough people at pain or higher, from more than one source, say so and one shows money. Public pages only, robots.txt and terms checked, GET only, and nobody in the community is ever contacted. Proposals wait for your yes at your own terminal; fifteen read-only checks run on the set.
Problem Mining Loop: five rules and a five-check script (free sample)
From the Problem Mining Loop: five rules for a complaint log you can trust (one owner and one buyer type, a link and date on every complaint, quote never paraphrase, money shown not guessed, count people not posts), and a read-only script that checks one log for them.
Evidence Standard: what counts as proof for each kind of agent work, who checks it, and what is refused
Agents report done from the plan as often as from the result. Five plain files write down, for each class of work, the proof that counts, who checks it (never the producer), what is refused and how long proof is kept. Agents log each claim with its evidence, seal it by sha256 and report in the words evidence.py say allows: done, partly done with the gap named, waiting on the owner, or reported, not verified. The agent's own summary is never proof. Fifteen read-only checks run on the standard and the claims ledger.
Evidence Standard: five rules and a five-check script (free sample)
From the Evidence Standard: five rules for what your agents may call done (a person owns the standard, every class of work names its proof and its checker, the agent's own account is never proof, people, money, records and releases are checked by a person, nobody says done on their own work), and a read-only script that checks one small file for them.
Model Provenance Record: where an open-weight model came from, how it behaves, and who said yes, before it is installed
Before any open-weight model goes on a business machine, fill one record: the publisher and exact revision, the license and whether it allows your use, every file with its published and local sha256, a header check that the format cannot run code (or a recorded scan if it can), the lineage back to the original, and a 15-test guardrail behaviour test that must still refuse clear misuse and still answer ordinary questions. Each result is tied to the hash of the weights and the test text, judged by someone who did not run it, and the owner says yes to one install hash. Fifteen read-only checks run on the record and the files.
Model Provenance Record: five rules and a five-check script (free sample)
From the Model Provenance Record: five rules before an open-weight model is installed (a named person says yes, pin the bytes, read the license first, no pickle without a scan, test behaviour on these exact weights), and a read-only script that checks one small file for them.
Source Shelf: a short, owner-approved list of industry reports your team reads, and a way to read them
Keep a small set of major industry reports as your team's default evidence base: each with its edition, publish date, link on the publisher's own site, the date someone last looked for a newer edition, and a note on what the publisher sells. Read any one of them against three questions (what is changing, who is struggling, what could help them), with word-for-word quotes and page numbers, the third answer marked as your inference, and every figure matched to its quote. Every add, drop and replace waits for the owner's yes. Fifteen read-only checks run on the folder.
Source Shelf: five rules and a five-check script (free sample)
From the Source Shelf: five rules for a small list of industry reports (one owner, link to the publisher not the post, dated and checked, three questions every time, every figure with its quote), and a read-only script that checks one small file for them.
Replicate Before You React: a trial kit for claims about AI agents
Test a claim about a coding agent on your own setup before you change anything: a security hole, a token-saving prompt, a hidden cost, a skill that makes the agent better. Restate the claim, run each setup 10 to 20 times in fresh containers with one capped key, use a harmless canary instead of a real payload, run a control arm for cost and quality, and get rates with honest intervals and a result card that says what you did not test.
Replicate Before You React: a claim card, five rules and a rate script (free sample)
From Replicate Before You React: a ten-line claim card to fill in before you test a claim about your agent, five safety rules (canary never payload, throwaway environment, one capped key, a control arm, say what you ran) and a script that shows how much a handful of runs can actually tell you.
Compounding Offer Audit: score an offer on six tests, find its weakest point, change one thing
Score each thing you sell from 0 to 3 on six tests: recurring, productized, systemized, owned demand, compounding, and whether growing it would make the owner's week better. Every score needs a checkable fact and its source, or the word unknown and the fact you are missing; no 3 on the owner's say-so, and no total. The audit names the weakest test, treats a low life score as a veto on growth, and proposes one change with a success test and a date. Every change waits for the owner's yes. Fifteen read-only checks run on the folder.
Compounding Offer Audit: five rules and a five-check script (free sample)
From the Compounding Offer Audit: five rules for scoring an offer on six tests (one owner, six scores and no total, a fact for every score, no 3 on say-so, one change on the weakest test), and a read-only script that checks one small file for them.
Agent-Ready Storefront Check: can an AI agent find, read and buy from your store, and which file fixes it
Fifteen read-only checks on the folder your build writes and the repo behind it, not a live URL: robots.txt and an export of your CDN or firewall bot rules let agent shoppers reach product and checkout pages, llms.txt links every machine-readable file, Product JSON-LD and prices agree, every paid page has a real checkout link, buy controls are links or buttons, no CAPTCHA sits on a product or checkout page, and policies are linked. Every finding names the source file to change. Source edits wait for the owner's yes; checkout and bot rule changes stay with the owner.
Agent-Ready Storefront Check: five rules and a five-check script (free sample)
From the Agent-Ready Storefront Check: five rules for a store agent shoppers can use (let the agents in, link every machine-readable file from llms.txt, Product JSON-LD with a price, a real checkout link on every paid page, no CAPTCHA on product pages), and a read-only script that checks your built site folder for them.
Startup Credits Sweep: an eligibility matrix, honest applications and an expiry ledger
Sweep startup programs for cloud, AI API and SaaS credits from your own facts. Each fact names its record; each program is read on its own page with the date; each requirement is marked meets, fails or unknown against one fact, and the matrix sets the verdict. Application drafts cite every fact and say it word for word, so nothing is stretched or invented to fit a form. A ledger counts credits only after the approval notice and reminds you before each grant expires. Every application waits for the owner's yes, and the owner submits it by hand. Fifteen read-only checks run on the folder.
Startup Credits Sweep: five rules and a five-check script (free sample)
From the Startup Credits Sweep: five rules for going after startup credits honestly (facts first, the program's own dated page, apply only when every requirement is met, answers say your facts word for word, count credits only after approval and set a reminder before expiry), and a read-only script that checks one small file for them.
Research Desk: five rules for research you will act on, and a five-check script (free sample)
From the Research bundle: five rules for research that feeds a decision (an owner and an auditor who did not do the research, a decision named for every question, one of five verdicts per claim, a date on every source, a primary source behind every CONFIRMED), and a read-only script that checks one small file for them.
AI Answer Visibility Check: does an assistant name you when a buyer asks, and who gets named instead
Ask each buyer question three times per assistant in fresh chats, paste the answers, and get a dated record of where you were named, where you were missed, who was named instead and which sources each answer cited. Fifteen read-only checks keep it honest: no brand in the question, three runs before a gap, answers hashed as given, nothing older than a month treated as current, and only your own brand. Page changes wait for the owner's yes.
AI Answer Visibility Check: five rules and a five-check script (free sample)
From the AI Answer Visibility Check: five rules for checking whether assistants name you (leave your name out of the question, a fresh chat each time, three runs before you call it, save the answer as it came, change your own pages with your own yes), and a read-only script that checks your saved answers.
Veteran GovCon AI Kit: capability statement, SAM.gov profile checklist and past performance with your agent
For veteran-owned and other small firms starting in federal contracting: a fact sheet your agent drafts from, eight prompts for a one-page capability statement, a SAM.gov profile checklist, a past performance skill, agency research prompts and capstat_check.py, thirteen read-only checks for UEI, CAGE and NAICS shape, hype words, placeholders, length and a dated source line for any status you list.
Capability statement ground rules and a five-line fact check (free sample)
From the Veteran GovCon AI Kit: the ground rules that keep an agent to your own facts when drafting a capability statement, and a five-line manual check for code shapes, contact details, hype words and dated status lines.