How to use the projects in your development environment
Last updated 6 November 2026
This guide is for developers who have seen the 6 projects and want to know what to do with them. For each project it gives what the project adds, its skill if it has one, a prompt to paste into your development assistant, and the steps; each project's own site has the details. Where the projects fit in a solution lifecycle explains when to use each one; this page explains how.
Before you start
- What you need. An AI assistant: a chat assistant (Claude, ChatGPT, Gemini, Copilot) for writing and review, and a development assistant that works in your repository (Claude Code, Codex, Antigravity CLI, Cursor, Copilot, or similar) for everything else. The AIQT and OPFiles tools and hooks need a recent Python 3, at the version their sites state; the other projects need no software of their own.
- 2 kinds of prompt. A prompt marked Published is the project's own text, quoted from the page named beside it: paste it unchanged. A prompt marked Composed for this guide is written here from the project's documentation, for a step the project does not publish a prompt for: adjust its bracketed parts to your project.
- Plan first, then approve. Every prompt below asks the assistant to propose a plan and to change nothing until you approve it. Read the plan, check the files it names, and approve or correct it. Keep the assistant's final report: it is the record of what was done and what could not be checked.
- When an assistant cannot open a link. Some assistants cannot fetch web pages. Download the file the prompt names and attach or paste it instead, and tell the assistant that you did.
The order to set them up
For a new or existing repository worked on with an AI assistant, this order lets each project build on the one before it.
| Step | Project | What it adds | When |
|---|---|---|---|
| 1 | AIQT Guardrails | The standard the assistant works to, and opt-in hooks that block or warn on specific actions | Once, at the start |
| 2 | OPFiles | Operational records (backlog, findings, decisions, worklog) with a drift check | Once, with AIQT or on its own |
| 3 | Clean Language | Writing rules for documentation, release notes and reports | Once, if the assistant writes prose people rely on |
| 4 | Secure Config | A guide per service, rules for every service and a verification checklist | At each deployment or change of a service |
| 5 | GRC Library | Policies, standards and registers to adopt, with framework mappings | When you need governance documents or a gap assessment |
| 6 | Audit This | A planned, evidence-based assessment of the project | Before a release, a handover, or an assurance request |
1. AIQT Guardrails
The standard your assistant works to: it checks a claim or says it did not, reports a shortcut, and calls work done only when it is. Details: aiqt.ai.
Skill: download it from aiqt.ai/install/
- Chat assistant: add the skill as aiqt.ai/install/ shows for your assistant.
- Development assistant: open it in your repository and paste the prompt below.
- Read the plan it proposes, then approve or correct it.
Published aiqt.ai/install/
Install AIQT Guardrails in this project and walk me through adopting it. Read https://aiqt.ai/llms.txt and the repository https://github.com/jposluns/guardrails first. Propose an adoption plan: the AIQT rules, hooks, and checks that fit this project's languages and tools, and OPFiles (https://opfiles.ai/llms.txt) for the project's operational records unless I opt out. Before building any new guardrail, check whether jposluns/guardrails already has one that fits. Name every file you would create or change, and why. Change nothing until I approve the plan.
2. OPFiles
Your project's backlog, findings, decisions and worklog as files in the repository, with a check that they stay in step. It is under active development: check opfiles.ai for what works today.
- Paste the prompt below into your development assistant. The AIQT prompt includes OPFiles unless you opted out.
- Approve the plan. The assistant checks the download against posluns.dev/hashes.txt before applying it.
- Ask it to make the drift check a required commit or CI check.
Published opfiles.ai/adopt/
Read https://opfiles.ai/llms.txt and https://opfiles.ai/adopt.md, then inspect this project and propose a plan to adopt OPFiles (OPF). The plan should name every file you would create, import, or retire, and why. Do not change any files until I have reviewed and approved the plan.
3. Clean Language
Writing rules for documentation, release notes and reports; your facts, code and quotations stay as given. Details: cleanlanguage.ai.
Skill: cleanlanguage.zip (latest release)
- Chat assistant: upload the zip without unzipping it, as cleanlanguage.ai/install/ shows for your assistant.
- Claude Code: unzip it into
.claude/skills/cleanlanguage/. Antigravity CLI: unzip it into.agents/skills/cleanlanguage/. - Any other development assistant: paste the prompt below.
Composed for this guide from cleanlanguage.ai and the skill's own scope
Read the Clean Language standard at https://raw.githubusercontent.com/jposluns/cleanlanguage/main/site/downloads/cleanlanguage.md. Apply it to the prose you write in this repository: the README and other documentation, release notes, pull request descriptions and commit message bodies. Do not apply it to code, commands, identifiers, configuration or quoted text, and keep them exactly as they are. Propose where to record this instruction so that you follow it in every session (for example this project's agent instruction file), and wait for my approval before adding it.
4. Secure Config
A guide per service for how it is exposed (TLS, authentication, MFA, secrets, network access), with a checklist to verify the result. Details: secureconfig.ai.
- Paste the prompt below into your development assistant. It audits first and lists the gaps.
- Approve the fixes you want; it then applies them and runs the checklist.
- At each deployment or change of a service, ask it to apply the matching guide from secureconfig.ai/llms.txt and run the checklist again.
Published secureconfig.ai/for-humans/
Secure this project using the guides at https://github.com/jposluns/secureconfig (see https://secureconfig.ai). Audit it first: list every listening service with its TLS and authentication state, and report the gaps as a table. Propose the fix for each gap per the matching guide (TLS on every endpoint, strong authentication, MFA where viable, and no secrets in the repository) and wait for my approval before changing anything. After approval, apply the fixes, run the README verification checklist, and report anything you could not test.
5. GRC Library
Policies, standards and registers to adopt for governance, risk, compliance, security, privacy and AI, with mappings to the common frameworks. Details: grclibrary.ai.
- Paste the prompt below into your development assistant.
- Approve the set of documents it proposes.
- Give it your roles and names to fill in; it keeps the licence notice the documents require.
Composed for this guide from the adopter guide and the licence
Read https://grclibrary.ai/llms.txt and the adopter guide at https://github.com/jposluns/grc_library/blob/main/docs/adopter-guide.md. For this project, propose the smallest set of documents to adopt from dev-security/ and, if this project builds or runs AI systems, ai/, with a one-line reason for each. Wait for my approval. Then copy the approved documents into [docs/governance/], replace the generic roles and placeholders with the values I give you, keep each document's metadata block, and keep the attribution and CC BY-SA 4.0 notice the licence requires, recording what you changed. Do not put real names or contact details in a public repository.
6. Audit This
A planned, evidence-based assessment of your project, in 2 steps: plan it, then run it. It changes nothing you have not approved. Details: auditthis.ai.
Skill: download it from auditthis.ai/install/
- Claude Code: unpack the skill so that the folder is
.claude/skills/auditthis/, then ask "Plan an assessment of this project with auditthis". Any other assistant: paste the prompt below. - Answer its intake questions, then review the plan.
- Run the audit with the audit prompt, and approve each fix by name.
Composed for this guide from auditthis.ai/for-ai/
Fetch https://auditthis.ai/prompts/plan.md and follow it exactly, starting with its intake questions. Do not read the project until I have answered them. If you cannot fetch the file, say so and stop.
What this guide does not claim
These projects help you assess, verify, find, remediate and demonstrate. None of them makes a solution or an organization compliant, secure or complete, and none replaces a certification, an attestation or legal advice. Each project's site states its current version, status, scope and limits, and those statements take precedence over this guide.