Record Upgrade Instructions logo

Record Upgrade Instructions

OrganizationPopular
prisma
record-upgrade-instructions

Record upgrade instructions alongside a Prisma 8 breaking-change PR, so downstream consumers (users of `@internal/*` and authors of Prisma 8 extensions) can apply the matching code translation automatically via the published upgrade skills. Use when you have refactored framework code and the test suite went red in `examples/` or `packages/3-extensions/`, when you fixed those red tests by updating example or extension code, when you are told to "record upgrade instructions for this PR", or when you made a breaking change to Prisma 8 that downstream consumers will need help migrating across.

Overview

Publisherprisma
Repositoryorm
Skill namerecord-upgrade-instructions
Stars
47.6K
Forks
2.5K
Bundled files
Instructions only
LicenseApache-2.0
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by prisma on GitHub. Read the source before you install it.

Installation

Install the Record Upgrade Instructions AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/prisma/orm.git /tmp/orm
mkdir -p .claude/skills
cp -r /tmp/orm/skills-contrib/record-upgrade-instructions .claude/skills/record-upgrade-instructions
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Record Upgrade Instructions in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Record Upgrade Instructions on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Record Upgrade Instructions is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Record upgrade instructions

Contribute an independent fragment on the PR that changes Prisma 8. Release preparation combines fragments into the published app and extension guides; feature PRs do not edit shared transition guides or choose a release number. Read the canonical upgrade instruction lifecycle for storage, release assembly, and checker modes.

Detection signals and routing

Use this skill when a framework change makes tests red in examples/ or packages/3-extensions/ and you fix them by updating the example or extension code rather than reverting the framework change. Those edits are the same translation downstream consumers need.

Directory changed by the PRFragment audienceConsumers
examples/appPublic package API, contract files, on-disk migrations
packages/3-extensions/extensionFramework SPI and extension authors
BothBoth, independentlyBoth audiences

Changes in these directories require a declaration, subject to existing coverage-check exclusions. Generated artefacts are not generally exempt: contract format changes require a codemod or re-emission instructions. Genuinely consumer-invisible changes still get an explicit changes: [] declaration, with no prose.

Stacked PRs: compare against the branch the PR actually targets, not always main. Each PR adds its own declaration in its own commits; inherited fragments do not cover a new PR. Do not pool a stack's instructions in its bottom PR.

Throughout this skill, <target> is the target branch (main unless stacked), <base> is its pinned comparison commit, and <head> is the PR head commit.

Authoring workflow

  1. Identify affected audiences. Inspect git diff <base>..<head> -- examples/ packages/3-extensions/. If neither directory has relevant changes, the coverage check does not require a declaration.

  2. Choose a descriptive pending name. Add upgrade-instructions/pending/<descriptive-name>/<audience>/instructions.md for each affected audience. Avoid collisions with pending work; no random suffix, global registry, historical-name reservation, or shared index is required. Never append to another PR's fragment. A release landing during your PR does not change this unversioned destination.

  3. Write the instructions. Retain the existing YAML frontmatter changes[] and Markdown prose format. Each change has a kebab-case id unique within the guide, a one-line summary, optional detection (glob and content predicate), and an optional script path relative to instructions.md. Prose-only transformations omit script; the consumer's agent follows the body.

    Make detection predicates token-precise. Test against both a true positive and the nearest false positive. A moved tag must not match an unchanged tag; excluding an unchanged spelling needs a token boundary:

    text
    matches:      .raw`
    must not fire on: fns.raw`
    too broad:    (?<!fns)\.raw`        also suppresses myfns.raw`
    shipped:      (?<!(?<![\w$])fns)\.raw`

    The inner lookbehind restricts the exclusion to the exact fns token.

    Only describe consumer action. Omit narrative about internal renames, dev-only dependency bumps, and incidental generated churn. If this PR's audience needs no action, use only:

    md
    ---
    changes: []
    ---

    No "consumers need not do anything" body prose. Both audiences declare independently, including real-change/no-op combinations.

  4. Author optional colocated scripts/assets. TypeScript (pnpm exec tsx), shell, or codemods are appropriate. Require no network, environment variables, or input beyond the consumer filesystem and bundled assets. Keep relative references inside the fragment's audience directory. For cross-audience changes, copy scripts into both audience directories; do not symlink or import from the other audience. The published clusters remain independently installable.

  5. Validate by execution using the unchanged concrete procedure below. An entry updates consumer code, not the example or extension tests. Do not introduce a separate testing system.

  6. Include the fragment and updated example or extension code in the PR. Commit working changes before running the Git-ref check:

    bash
    pnpm check:upgrade-coverage --mode pr --prev <base> --head <head>

    <head> must include the committed declaration; the check does not inspect uncommitted edits. Link each new fragment directory in the PR description. Review must verify that the instructions actually describe the PR's changes; the gate checks added declarations, readable change lists, and relative script references, not semantic correctness.

Validation by execution

Before merging, run every new entry against the corresponding example or extension code in this repository, starting from its pre-PR state and ending with passing tests. This is the existing per-PR quality bar, not a release-wide migration test.

Workflow per entry (both flows apply for cross-audience entries):

  • Open PR: <head> is the PR branch head; <base> is the actual target branch's comparison commit.
  • Merged PR: <head> is the merge commit; <base> is its mainline parent. git log --first-parent names both.

Use a disposable checkout for the restoration steps so unrelated working changes are not overwritten. The PR author updates example and extension tests; the upgrade instructions must not change them. The equality check excludes test/ directories; the companion check confirms the entry left those directories exactly as it found them.

App entry (against examples/)

  1. Check out <head>, which has the framework change applied.

  2. Revert examples/ to its pre-PR state (git restore --source=<base> -- examples/).

  3. Run the fragment against the restored example code: invoke colocated scripts per script: references, then walk the prose body for additional instructions.

  4. Verify examples/ matches <head> outside test directories:

    bash
    git status --porcelain -- examples/ ':(exclude)examples/*/test/**'

    It must print nothing, including no newly created files. The entry must reproduce git diff <base>..<head> -- examples/ ':(exclude)examples/*/test/**'.

  5. Verify the entry left tests at <base>; the source equality check cannot detect test changes:

    bash
    git diff --exit-code <base> -- 'examples/*/test/**'
    git ls-files --others --exclude-standard -- 'examples/*/test/**'

    The first command must exit 0 and the second print nothing. An entry must not mutate or create tests to make the next step pass.

  6. Run pnpm --filter <example-package> test for each touched example. The repo-wide pnpm test:examples also runs examples needing a database and .env (pnpm db:up, then copy .env.example); run it only with those in place.

Extension entry (against packages/3-extensions/)

  1. Check out <head>, which has the framework change applied.

  2. Revert packages/3-extensions/ to its pre-PR state (git restore --source=<base> -- packages/3-extensions/).

  3. Run the fragment against the restored extension code, including referenced scripts and prose.

  4. Verify the non-test paths match <head>:

    bash
    git status --porcelain -- packages/3-extensions/ ':(exclude)packages/3-extensions/*/test/**'

    It must print nothing. The entry reproduces git diff <base>..<head> -- packages/3-extensions/ ':(exclude)packages/3-extensions/*/test/**'.

  5. Verify test paths remain at <base>:

    bash
    git diff --exit-code <base> -- 'packages/3-extensions/*/test/**'
    git ls-files --others --exclude-standard -- 'packages/3-extensions/*/test/**'

    The first command must exit 0; the second must print nothing.

  6. Verify the matching test suite is green: pnpm test --filter='./packages/3-extensions/*'.

If any check fails, iterate on the entry; do not merge. Classify failures before changing anything, per CI failure classification. A timeout or connection error makes the environment a candidate cause, not a verdict.

PR commit shape

Include:

  • Each new upgrade-instructions/pending/<name>/<audience>/instructions.md and any colocated scripts/assets.
  • The updated example and extension code, matching the result of applying the instructions outside test directories. The entry neither writes nor updates those tests.
  • PR-description references naming the fragment directories, for example upgrade-instructions/pending/migration-metadata-shape/app/ and upgrade-instructions/pending/migration-metadata-shape/extension/.

Both audience copies may share IDs, summaries, or detection predicates; they are independent records. Fixes to either copy use normal PR review. Historical published guidance can also be corrected through normal reviewed PRs, but edits to old guides do not replace a new PR's required pending declaration.

Out of scope

Fragments describe code translation only. Do not add the general bump/install/instructions/validate/commit loop to their bodies: the published app and extension flows own it. Extension exact-pin enforcement remains prisma-8-check-pins from @internal/extension-author-tools.

Release synthesis, archives, skipped unpublished bumps, and release completeness belong to the canonical lifecycle, not feature-PR authoring. No release-wide migration rehearsal is required.

Frequently asked questions

What does the Record Upgrade Instructions AI skill do?

Record upgrade instructions alongside a Prisma 8 breaking-change PR, so downstream consumers (users of `@internal/*` and authors of Prisma 8 extensions) can apply the matching code translation automatically via the published upgrade skills. Use when you have refactored framework code and the test suite went red in `examples/` or `packages/3-extensions/`, when you fixed those red tests by updating example or extension code, when you are told to "record upgrade instructions for this PR", or when you made a breaking change to Prisma 8 that downstream consumers will need help migrating across.

Why use Record Upgrade Instructions on TypingMind?

Because you install it once and use it with any model. Record Upgrade Instructions is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Record Upgrade Instructions in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/prisma/orm/tree/main/skills-contrib/record-upgrade-instructions. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Record Upgrade Instructions?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Record Upgrade Instructions?

As many as you like. As long as a model supports skills, you can use Record Upgrade Instructions with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Record Upgrade Instructions AI skill free?

Yes. It is published on GitHub by prisma under the Apache-2.0 license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇