Vibe GuidePrepare deployment or packaging
Installers, static pages and backend-powered sites need different preparation.
Why this milestone?
Prepare the chosen files and actually try them using the instructions.
Start here2 steps · Follow in order
0102Prepare and try the deliverableTo checkOpen this stepClose this step
01What do you need, and what should you keep?
Your checked delivery preparation list.
The next step can find the same deliverable.
Which files should AI read and save in this step?
In your tool’s project chat, ask it to open the files below. If a file is missing, follow the link back to the step that creates it.
Ask AI to read these materials first: docs/design.md · docs/delivery.md · docs/release.md
Ask AI to save the result to: docs/release.md
Missing these materials? Return to the step that creates them →
02What will this step help you do?
Prepare the chosen files and actually try them using the instructions.
Understand it, then judge for yourself
What does this concept mean?
Installers, static pages and backend-powered sites need different preparation.
How would you decide in a different situation?
If a package has not run on the target OS, is installation verified?
See a reference answer
Record build success separately from an unverified target trial.
Answer in your own words; if unsure, compare the explanation above and check it against your own project.
03What exactly should you do in this step?
OpenClose
Project chat and a test environment
Send one action at a time: select and copy its prompt (⌘C on Mac, Ctrl+C on Windows), switch to your tool’s project chat, paste it, replace bracketed fields and send. Wait for the reply and check the stated result before continuing.
Prepare files for the chosen delivery method.OpenClose
Suggested prompt:
Read docs/design.md, docs/delivery.md and docs/release.md. Prepare the agreed files and explain location, dependencies and startup. Local use may use source and instructions without an installer.
Afterwards, you should see: Real files and matching instructions are available.
Try it independently as the recipient would.OpenClose
Suggested prompt:
Give an isolated trial method without relying on an existing development service. Test opening, saving, reading and reopening with test data while preserving the original.
Afterwards, you should see: The instructions let the recipient use core features.
Save the trial record.OpenClose
Suggested prompt:
Write artifact location, revision, real entry, trial results and recovery to docs/release.md. Keep untested items and do not publish yet.
Afterwards, you should see: The next step can find the same deliverable.
Practice: personal journal (frontend + backend + database)
Use the same learning-journal project throughout this example. For your own project, compare applicable requirements without creating another project.
Prepare startup instructions/dependencies and trial an isolated copy with test data rather than private records.
- What should you see after this step?
- A clean copy starts and saves/reads one record.
- What should you do first if you get stuck?
- Do not treat a copied .venv as a portable package; recreate the environment from dependencies.
04Not sure how? Follow each action
OpenClose
Prepare files for the chosen delivery method.
- What should you see after this action?
- Real files and matching instructions are available.
- What should you do if that result is missing?
- Use configuration placeholders; exclude real secrets and user databases from public packages.
Try it independently as the recipient would.
- What should you see after this action?
- The instructions let the recipient use core features.
- What should you do if that result is missing?
- Compare dependencies/configuration/paths if it only works in development; mark unavailable devices untested.
Save the trial record.
- What should you see after this action?
- The next step can find the same deliverable.
- What should you do if that result is missing?
- Fix failed trials rather than equating file generation with usability.
05Words in this step
OpenClose
Build ↗
Processes source files into runnable or distributable artifacts.
Manufacture a product, then still inspect it.
Artifact ↗
Files or packages produced by a build.
The actual manufactured output.
Code signing ↗
Adds verifiable publisher and integrity information to software.
A seal verifies origin, not absence of defects.
06How should you handle a problem here?
OpenClose
Installer or deployed application fails
- Where should you look to understand the problem?
- Check target OS, build version and exact errors; do not bypass OS security warnings.
- Follow these steps to resolve the problem
- Keep the artifact and logs, return to feedback, and check compatibility, dependencies and legitimate signing/platform requirements. Retest without replacing a working version.
- Expected recovery
- Repeat the original action and obtain the expected result; keep unexecuted checks marked unverified.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
I am at “Prepare and try the deliverable”. Expected: The deliverable has been tried, with use and recovery instructions; release still needs personal approval.. Actual screen/error: [add here]. Inspect existing files first. Give one reversible action at a time, its location, and expected result. Do not recreate the project or delete existing work.07Material for this step (fill or copy)
OpenClose
If you completed the individual prompts above, do not resend this full template. Use it to organize this step’s material when needed; replace bracketed fields, confirm and copy into the project chat.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
First check access to docs/design.md, docs/delivery.md, docs/release.md. If missing or inaccessible, identify the prerequisite and owner action, then stop rather than invent prior work.
Read approved docs/release.md. Prepare the appropriate build/deployment configuration or installer. Document prerequisites, execution location, costs, configuration, success and recovery. Keep secrets out of source and packages. Actually trial deploy/install and check core actions, reopening and persistence. Record version/results and unverified work. Do not publish without approval or add Docker by default. Check data retention only if required; record actual artifact location, dependencies and evidence in docs/release.md.Journal practice: fill or supplement this step
For this full-stack practice only: use this combined prompt, fill its fields and send it once instead of also sending the generic template. It does not authorize later steps.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
First check access to docs/design.md, docs/delivery.md, docs/release.md. If missing or inaccessible, identify the prerequisite and owner action, then stop rather than invent prior work.
Read approved docs/release.md. Prepare the appropriate build/deployment configuration or installer. Document prerequisites, execution location, costs, configuration, success and recovery. Keep secrets out of source and packages. Actually trial deploy/install and check core actions, reopening and persistence. Record version/results and unverified work. Do not publish without approval or add Docker by default. Check data retention only if required; record actual artifact location, dependencies and evidence in docs/release.md.
I selected the full-stack journal practice.
Prepare local delivery only: directory, OS-specific environment/dependencies, non-destructive database setup, start/stop, actual URL and backup/restore. Test an isolated copy, name untested systems and do not publish.
Identify the isolated delivery test location; do not rely on an existing development service. Trial entry, save, read and restart; record URL, revision, dependencies and untested work in docs/release.md. Use configuration placeholders without real keys or user databases.08Check this before continuing
The deliverable has been tried, with use and recovery instructions; release still needs personal approval.
09What should you do if the expected result is missing?
OpenClose
Use configuration placeholders; exclude real secrets and user databases from public packages.
Compare with an example (illustration, not your result)
Repository → actual build command → output/installer → test target → user actions → evidence.
Expand the instructions for my project
Choose only relevant branches. Complete one action at a time using its prompt; do not resend the full template after completing individual actions. AI derives commands/config from the actual project; verify current official account/channel requirements.
Personal local use
When no public URL, store or third-party installation is needed.
Open README and follow the startup instructions.
Suggested prompt:
I am working on Personal local use. Do only this action: Open README and follow the startup instructions. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
If you want double-click startup, ask the tool to prepare an entry for your computer.
Suggested prompt:
I am working on Personal local use. Do only this action: If you want double-click startup, ask the tool to prepare an entry for your computer. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Locate saved data, make a backup and test restoration using a copy.
Suggested prompt:
I am working on Personal local use. Do only this action: Locate saved data, make a backup and test restoration using a copy. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
How can you check that this branch is complete?You can reopen it as documented and locate/recover relevant data.
Prompt for this branch
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
Read docs/design.md, docs/delivery.md, docs/release.md and confirm this branch is in agreed scope. Stop on missing files rather than inventing a plan.
Read the local-use decision and README. Prepare only repeatable local delivery: actual entry, dependencies and target OS. Explain any launcher/package before creating it, without adding cloud services. Verify restart and required persistence; record real paths, revision, backup/recovery and untested work in docs/release.md. Do not push, publish or widen access.Web: static output
For HTML/assets without a backend process.
Ask whether the site only needs static webpage files.
Suggested prompt:
I am working on Web: static output. Do only this action: Ask whether the site only needs static webpage files. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Ask the tool to build the deliverable files and explain how to preview them.
Suggested prompt:
I am working on Web: static output. Do only this action: Ask the tool to build the deliverable files and explain how to preview them. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Create a test site on the chosen host, enter the required settings and try its temporary URL.
Suggested prompt:
I am working on Web: static output. Do only this action: Create a test site on the chosen host, enter the required settings and try its temporary URL. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
How can you check that this branch is complete?The real build works through the test URL, including assets and reloads.
Prompt for this branch
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
Read docs/design.md, docs/delivery.md, docs/release.md and confirm this branch is in agreed scope. Stop on missing files rather than inventing a plan.
Read design, delivery, release and build config. Prepare only the selected static site. Confirm no required server process, report commands/directory/output/config names, build and test assets, deep-link reloads and user flows in an available test environment. Explain owner console actions and success signals without requesting secrets. Record artifact, revision, test URL, results and recovery in docs/release.md. Do not release publicly; return to delivery selection if not static.Web: frontend plus external services
When hosting depends on external data, login or APIs.
Ask which external services and accounts the site needs.
Suggested prompt:
I am working on Web: frontend plus external services. Do only this action: Ask which external services and accounts the site needs. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Open each service dashboard and enter the site URL and connection settings where instructed.
Suggested prompt:
I am working on Web: frontend plus external services. Do only this action: Open each service dashboard and enter the site URL and connection settings where instructed. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Use an ordinary test account to try login, saving and reading; check costs and failure feedback.
Suggested prompt:
I am working on Web: frontend plus external services. Do only this action: Use an ordinary test account to try login, saving and reading; check costs and failure feedback. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
How can you check that this branch is complete?The target works without the development machine and with correct configuration/access.
Prompt for this branch
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
Read docs/design.md, docs/delivery.md, docs/release.md and confirm this branch is in agreed scope. Stop on missing files rather than inventing a plan.
Prepare only the selected frontend and external-service trial. List test/production environments, config names, owner setup, callbacks/origins/access requirements without secret values. Verify real login, data and API behavior through the target URL, including relevant rejection and quotas. Record dependencies, billing entry, backup/recovery, revision and evidence. Mark missing prerequisites blocked; do not release or provision paid resources.Web: server-rendered or backend app
When server code, jobs or databases must run.
Ask what the server needs to run and its estimated cost.
Suggested prompt:
I am working on Web: server-rendered or backend app. Do only this action: Ask what the server needs to run and its estimated cost. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
After choosing a test server, request one step at a time for uploading, configuring and starting.
Suggested prompt:
I am working on Web: server-rendered or backend app. Do only this action: After choosing a test server, request one step at a time for uploading, configuring and starting. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Complete a task at the test URL, then restart the service and check retained data.
Suggested prompt:
I am working on Web: server-rendered or backend app. Do only this action: Complete a task at the test URL, then restart the service and check retained data. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
How can you check that this branch is complete?The target runs independently with required persistence and verified flows/recovery.
Prompt for this branch
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
Read docs/design.md, docs/delivery.md, docs/release.md and confirm this branch is in agreed scope. Stop on missing files rather than inventing a plan.
Prepare the actual server trial: runtime/dependencies, artifacts, config names, migrations, start command, ports/proxy, HTTPS, logs and restarts. Confirm access/costs; back up and test migrations in a test database. Execute only authorized test actions, verify ordinary-user flows and persistence, and record evidence and compatible recovery in docs/release.md. Do not assume Docker or change production data/release.Android: direct download or store
For Android; choose a distribution channel first.
Choose the Android devices and direct download or store distribution.
Suggested prompt:
I am working on Android: direct download or store. Do only this action: Choose the Android devices and direct download or store distribution. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Ask for release files for that channel and instructions for storing signing keys.
Suggested prompt:
I am working on Android: direct download or store. Do only this action: Ask for release files for that channel and instructions for storing signing keys. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Install on the target phone and check that an upgrade retains existing data.
Suggested prompt:
I am working on Android: direct download or store. Do only this action: Install on the target phone and check that an upgrade retains existing data. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
How can you check that this branch is complete?Channel-specific output works on real devices; review-pending is not live.
Prompt for this branch
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
Read docs/design.md, docs/delivery.md, docs/release.md and confirm this branch is in agreed scope. Stop on missing files rather than inventing a plan.
Read the Android channel decision and current official requirements for identity, version, signing and developer verification. Explain owner account/key setup without exposing secrets. Build actual channel-appropriate release output and distinguish APK from upload artifacts. Test installation, permissions, core flows and update retention on available target devices; mark missing tests untested. Record channel, revision, output and evidence; do not submit or publish downloads.iPhone/iPad: testing and distribution
For iOS/iPadOS, not Android distribution.
Ask for required computer, developer account and devices, then check what you have.
Suggested prompt:
I am working on iPhone/iPad: testing and distribution. Do only this action: Ask for required computer, developer account and devices, then check what you have. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Configure account and signing using current Apple instructions, then prepare a test version.
Suggested prompt:
I am working on iPhone/iPad: testing and distribution. Do only this action: Configure account and signing using current Apple instructions, then prepare a test version. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Install and try it on your iPhone or iPad; check store materials before public release.
Suggested prompt:
I am working on iPhone/iPad: testing and distribution. Do only this action: Install and try it on your iPhone or iPad; check store materials before public release. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
How can you check that this branch is complete?An eligible channel is identified, with device tests and review state separately recorded.
Prompt for this branch
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
Read docs/design.md, docs/delivery.md, docs/release.md and confirm this branch is in agreed scope. Stop on missing files rather than inventing a plan.
Verify current Apple requirements for devices, Xcode, account, Bundle ID, signing and the chosen channel. Prepare only applicable archive/upload/test actions; the owner handles credentials. Report actual output/version and test core flows, permissions and upgrades on available devices. Distinguish build, processing, testing, review and live states. Save evidence/untested items to docs/release.md; do not submit for review or assume hosted IPA installation.Alternative example: Apple web-distribution eligibility (not universal)
Desktop: Windows / macOS / Linux
For installed desktop applications.
Specify Windows, macOS or Linux and the computer model.
Suggested prompt:
I am working on Desktop: Windows / macOS / Linux. Do only this action: Specify Windows, macOS or Linux and the computer model. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Ask for the installer for that system and its signing/distribution requirements.
Suggested prompt:
I am working on Desktop: Windows / macOS / Linux. Do only this action: Ask for the installer for that system and its signing/distribution requirements. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Try installation, startup and upgrade on the target computer and check retained data.
Suggested prompt:
I am working on Desktop: Windows / macOS / Linux. Do only this action: Try installation, startup and upgrade on the target computer and check retained data. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
How can you check that this branch is complete?Ordinary target computers can install/use it and retain required data across updates.
Prompt for this branch
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
Read docs/design.md, docs/delivery.md, docs/release.md and confirm this branch is in agreed scope. Stop on missing files rather than inventing a plan.
Read desktop delivery decisions and verify OS/architecture, toolchain, channel and current signing requirements. Prepare release artifacts and installation/uninstall/data/update/recovery instructions. Test on available targets without development tools; mark missing devices unverified. Save evidence to docs/release.md; do not publish or bypass OS protections.Mini-program: chosen platform
For platforms such as WeChat; do not mix platform procedures.
Check registration and release eligibility on the chosen mini-app platform’s official site.
Suggested prompt:
I am working on Mini-program: chosen platform. Do only this action: Check registration and release eligibility on the chosen mini-app platform’s official site. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Open the project in the platform’s developer tool and check its ID and service URLs.
Suggested prompt:
I am working on Mini-program: chosen platform. Do only this action: Open the project in the platform’s developer tool and check its ID and service URLs. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
Prepare a trial build, test the full flow with an authorized phone account, then prepare review materials.
Suggested prompt:
I am working on Mini-program: chosen platform. Do only this action: Prepare a trial build, test the full flow with an authorized phone account, then prepare review materials. Inspect my project and explain where to act, the action and its expected result. For account actions, guide one action at a time and wait for me. Do not advance to another action or publish.
How can you check that this branch is complete?Trial works for real target accounts, prerequisites/material are complete, and unreleased stays unreleased.
Prompt for this branch
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
Read docs/design.md, docs/delivery.md, docs/release.md and confirm this branch is in agreed scope. Stop on missing files rather than inventing a plan.
Read the selected mini-program platform. Check current official account/category, app identity, developer permissions, request configuration and trial prerequisites. Explain owner console actions and success signals; create and test an actual trial on available devices. Record platform, revision, config names, evidence and untested items. Do not substitute local bypasses, use another platform’s procedure or submit/release without authorization.Check your account eligibility and review requirements in the official console, then prepare the requested materials.
Record this check and project version
This is self-reported, not automatic inspection. Old checks do not prove a changed version; complete or report failed and untested work.
0202Optional: publish a simple page on GitHub PagesOptionalTo checkOpen this stepClose this step
Optional · static sites only, with no private data and no server; prepare a plan, then return to this milestone
01What do you need, and what should you keep?
Your project files and agreed technology choices.
The preparation instructions are actionable.
Which files should AI read and save in this step?
In your tool’s project chat, ask it to open the files below. If a file is missing, follow the link back to the step that creates it.
Ask AI to read these materials first: docs/delivery.md
Ask AI to save the result to: Read-only; no file changes
Missing these materials? Return to the step that creates them →
02What will this step help you do?
Check whether your page fits GitHub Pages before proceeding.
Understand it, then judge for yourself
What does this concept mean?
Git records versions, GitHub hosts/shares code, and GitHub Pages hosts static webpages.
How would you decide in a different situation?
Why might local records differ at a production URL?
See a reference answer
Browser local data is origin-scoped. A new domain may require explicit backup/migration; it does not appear automatically.
Answer in your own words; if unsure, compare the explanation above and check it against your own project.
03What exactly should you do in this step?
OpenClose
Only static pages without private material or a server
Send one action at a time: select and copy its prompt (⌘C on Mac, Ctrl+C on Windows), switch to your tool’s project chat, paste it, replace bracketed fields and send. Wait for the reply and check the stated result before continuing.
Check whether GitHub Pages fits. It hosts static pages, not your backend/database service.OpenClose
Suggested prompt:
Read docs/delivery.md and assess static export suitability. Explain server/private-configuration dependencies and return to delivery selection if unsuitable. Assess only; do not publish.
Afterwards, you should see: Suitability is clear.
If suitable, prepare exact instructions and return to delivery preparation.OpenClose
Suggested prompt:
List this project’s static build, output folder and GitHub Pages settings with locations. Provide a plan only; do not create a remote repository, push or publish.
Afterwards, you should see: The preparation instructions are actionable.
Practice: personal journal (frontend + backend + database)
Use the same learning-journal project throughout this example. For your own project, compare applicable requirements without creating another project.
This optional action is for static sites. Continue to maintenance for this full-stack practice; do not upload Python/database to Pages.
- What should you see after this step?
- Pages cannot run this backend and the database stays private.
- What should you do first if you get stuck?
- Publishing just the page breaks saves; choose a delivery plan that runs the backend and retains data.
04Not sure how? Follow each action
OpenClose
Check whether GitHub Pages fits. It hosts static pages, not your backend/database service.
- What should you see after this action?
- Suitability is clear.
- What should you do if that result is missing?
- The full-stack journal cannot be fully deployed by uploading its page alone.
If suitable, prepare exact instructions and return to delivery preparation.
- What should you see after this action?
- The preparation instructions are actionable.
- What should you do if that result is missing?
- Review files for secrets/private content before approving public access.
05Words in this step
OpenClose
Static site ↗
A site primarily serving prebuilt files.
Hand readers a preprinted brochure.
Static site ↗
A site primarily serving prebuilt files.
Hand readers a preprinted brochure.
06How should you handle a problem here?
OpenClose
The public URL returns 404
- Where should you look to understand the problem?
- Check that Pages reports success and index.html exists at the selected branch root.
- Follow these steps to resolve the problem
- Check deployment status and the actual Visit site URL. If pending, wait for completion; if failed, retain logs for AI. Do not guess the website from the repository URL.
- Expected recovery
- The site URL shown by GitHub opens the intended page.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
I am at “Optional: publish a simple page on GitHub Pages”. Expected: A reviewed suitability decision and settings plan exist; preparation does not count as publication.. Actual screen/error: [add here]. Inspect existing files first. Give one reversible action at a time, its location, and expected result. Do not recreate the project or delete existing work.07Material for this step (fill or copy)
OpenClose
If you completed the individual prompts above, do not resend this full template. Use it to organize this step’s material when needed; replace bracketed fields, confirm and copy into the project chat.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
First check access to docs/delivery.md. If missing or inaccessible, identify the prerequisite and owner action, then stop rather than invent prior work.
Read project type, build method and docs/delivery.md. Review docs/release.md if it exists; otherwise mark delivery preparation pending without blocking this suitability assessment. Assess only GitHub Pages suitability: static output without a server runtime or private server credentials. If unsuitable, explain and return to delivery selection. Otherwise provide the actual build command, output directory and Pages settings for review. Prepare instructions, return to “Prepare deployment or packaging” for the trial run, and execute remote creation, push and release only after confirmation in “Deliver and check the real entry”.Journal practice: fill or supplement this step
For this full-stack practice only: use this combined prompt, fill its fields and send it once instead of also sending the generic template. It does not authorize later steps.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
First check access to docs/delivery.md. If missing or inaccessible, identify the prerequisite and owner action, then stop rather than invent prior work.
Read project type, build method and docs/delivery.md. Review docs/release.md if it exists; otherwise mark delivery preparation pending without blocking this suitability assessment. Assess only GitHub Pages suitability: static output without a server runtime or private server credentials. If unsuitable, explain and return to delivery selection. Otherwise provide the actual build command, output directory and Pages settings for review. Prepare instructions, return to “Prepare deployment or packaging” for the trial run, and execute remote creation, push and release only after confirmation in “Deliver and check the real entry”.
I selected the full-stack journal practice.
Assess suitability only: Flask/SQLite cannot be fully delivered by static Pages. Explain and return to delivery planning without publishing.08Check this before continuing
A reviewed suitability decision and settings plan exist; preparation does not count as publication.
09What should you do if the expected result is missing?
OpenClose
The full-stack journal cannot be fully deployed by uploading its page alone.
Compare with an example (illustration, not your result)
Evaluate brochure/docs sites; never place private API keys in browser output.
Examples and reference
Approve public files first. Do not upload secrets, chat logs, or personal journal data. The new URL uses separate local storage.
Not passed? Repair before continuingWhen testing or release review finds a problem, repair in three steps, then return to the milestone above.
0103Record the trial feedbackTo checkOpen this stepClose this step
01What do you need, and what should you keep?
The actions, screen or error from the failure.
Issues have IDs, reproduction steps and evidence.
Which files should AI read and save in this step?
In your tool’s project chat, ask it to open the files below. If a file is missing, follow the link back to the step that creates it.
Ask AI to read these materials first: Confirmed material and project location
Ask AI to save the result to: docs/feedback.md
Missing these materials? Return to the step that creates them →
02What will this step help you do?
Describe the failure precisely so the tool can find the same problem.
Understand it, then judge for yourself
What does this concept mean?
Feedback records actions and observations; you do not need to diagnose the code.
How would you decide in a different situation?
Which is an observation and which needs evidence: “button does nothing” or “database is broken”?
See a reference answer
An unresponsive button is an observation; a broken database is a hypothesis needing evidence.
Answer in your own words; if unsure, compare the explanation above and check it against your own project.
03What exactly should you do in this step?
OpenClose
The work, this template and project chat
Send one action at a time: select and copy its prompt (⌘C on Mac, Ctrl+C on Windows), switch to your tool’s project chat, paste it, replace bracketed fields and send. Wait for the reply and check the stated result before continuing.
Choose one issue and record the entry, input, clicks and result.OpenClose
Afterwards, you should see: Someone else can follow the same steps.
Fill the template with actions/errors, confirm and send it.OpenClose
Suggested prompt:
Organize these issues in docs/feedback.md with IDs, expected and actual results: 【paste record】. Record and investigate without changing code. Do not present guesses as facts.
Afterwards, you should see: Issues have IDs, reproduction steps and evidence.
Practice: personal journal (frontend + backend + database)
Use the same learning-journal project throughout this example. For your own project, compare applicable requirements without creating another project.
Collect test input, click sequence, time, page message and error together; omit private data from screenshots.
- What should you see after this step?
- Someone can reproduce the issue, rather than merely reading “save broke”.
- What should you do first if you get stuck?
- If logs are unfamiliar, ask Codex to locate this service’s output and extract the matching error.
04Not sure how? Follow each action
OpenClose
Choose one issue and record the entry, input, clicks and result.
- What should you see after this action?
- Someone else can follow the same steps.
- What should you do if that result is missing?
- Describe the symptom, not merely “broken”; you need not know the cause.
Fill the template with actions/errors, confirm and send it.
- What should you see after this action?
- Issues have IDs, reproduction steps and evidence.
- What should you do if that result is missing?
- Hide passwords/private information in screenshots before sharing.
05Words in this step
OpenClose
06How should you handle a problem here?
OpenClose
The issue happens intermittently
- Where should you look to understand the problem?
- Record occurrence count, device, version and the latest full sequence.
- Follow these steps to resolve the problem
- Collect symptoms, actions, revision and exact errors; use read-only diagnosis if needed. Do not repair here. Record feedback and confirm the plan in “Plan the fix and its passing checks” first.
- Expected recovery
- Verifiable records exist; untested and failed results keep their status.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
I am at “Record the trial feedback”. Expected: Feedback is reproducible and separates facts, guesses and new requests.. Actual screen/error: [add here]. Inspect existing files first. Give one reversible action at a time, its location, and expected result. Do not recreate the project or delete existing work.07Material for this step (fill or copy)
OpenClose
If you completed the individual prompts above, do not resend this full template. Use it to organize this step’s material when needed; replace bracketed fields, confirm and copy into the project chat.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
First check access to the current project location. If missing or inaccessible, identify the prerequisite and owner action, then stop rather than invent prior work.
Version/date: [fill in]
Device and entry: [fill in]
Feature: [fill in]
Steps: [fill in]
Expected: [fill in]
Actual: [fill in]
Evidence: [attach or none]
Does the problem happen every time or only sometimes: [always/sometimes/once]
Which action does this problem prevent you from completing: [blocked/workaround/cosmetic]
Passed and untested: [fill in]
Save these personal observations in docs/feedback.md with issue IDs. Preserve facts and unknowns; ask for missing material. Do not fix code yet.Journal practice: fill or supplement this step
For this full-stack practice only: use this combined prompt, fill its fields and send it once instead of also sending the generic template. It does not authorize later steps.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
First check access to the current project location. If missing or inaccessible, identify the prerequisite and owner action, then stop rather than invent prior work.
Version/date: [fill in]
Device and entry: [fill in]
Feature: [fill in]
Steps: [fill in]
Expected: [fill in]
Actual: [fill in]
Evidence: [attach or none]
Does the problem happen every time or only sometimes: [always/sometimes/once]
Which action does this problem prevent you from completing: [blocked/workaround/cosmetic]
Passed and untested: [fill in]
Save these personal observations in docs/feedback.md with issue IDs. Preserve facts and unknowns; ask for missing material. Do not fix code yet.
I selected the full-stack journal practice.
Name the failing check: table/API/connection/read/restart. Distinguish page messages from backend errors; do not paste full database contents or secrets.08Check this before continuing
Feedback is reproducible and separates facts, guesses and new requests.
09What should you do if the expected result is missing?
OpenClose
Describe the symptom, not merely “broken”; you need not know the cause.
Compare with an example (illustration, not your result)
F01 / cancel then refresh / expect one place restored / actual unchanged / revision abc / reproduced twice.
0203Plan the fix and its passing checksTo checkOpen this stepClose this step
01What do you need, and what should you keep?
The feedback record you just saved.
The fix and its test are specific.
Which files should AI read and save in this step?
In your tool’s project chat, ask it to open the files below. If a file is missing, follow the link back to the step that creates it.
Ask AI to read these materials first: docs/feedback.md · docs/requirements.md
Ask AI to save the result to: docs/repair-plan.md
Missing these materials? Return to the step that creates them →
02What will this step help you do?
Understand the proposed fix before letting the tool make it.
Understand it, then judge for yourself
What does this concept mean?
A repair plan explains changes, possible effects and retesting.
How would you decide in a different situation?
Why check history retrieval after fixing the Save button?
See a reference answer
Saving and reading may share structures/APIs; retest the fix and regress related behavior.
Answer in your own words; if unsure, compare the explanation above and check it against your own project.
03What exactly should you do in this step?
OpenClose
Project chat and docs/repair-plan.md
Send one action at a time: select and copy its prompt (⌘C on Mac, Ctrl+C on Windows), switch to your tool’s project chat, paste it, replace bracketed fields and send. Wait for the reply and check the stated result before continuing.
Ask why the issue occurs before changing anything.OpenClose
Suggested prompt:
Read docs/feedback.md and the requirements. Investigate without changing code. Separate facts from hypotheses and give the next small check.
Afterwards, you should see: The explanation fits the observed issue.
Review where the fix will happen and how you will check it.OpenClose
Suggested prompt:
Save docs/repair-plan.md with changed files, affected features, original-case retest and expected results. Include backup/recovery before database changes. Wait for my approval.
Afterwards, you should see: The fix and its test are specific.
Practice: personal journal (frontend + backend + database)
Use the same learning-journal project throughout this example. For your own project, compare applicable requirements without creating another project.
Ask Codex to reproduce and identify the layer to change and the failed action to rerun.
- What should you see after this step?
- The plan protects data and retests previously working saves.
- What should you do first if you get stuck?
- Stop a suggested database deletion; request backup and a non-destructive plan or more diagnosis.
04Not sure how? Follow each action
OpenClose
Ask why the issue occurs before changing anything.
- What should you see after this action?
- The explanation fits the observed issue.
- What should you do if that result is missing?
- Investigate further rather than defaulting to reinstalling or deleting storage.
Review where the fix will happen and how you will check it.
- What should you see after this action?
- The fix and its test are specific.
- What should you do if that result is missing?
- Ask for specifics if it only says optimize.
05Words in this step
OpenClose
Root cause ↗
The underlying cause rather than a symptom.
A dark lamp may be caused by wiring.
Regression testing ↗
Checks whether changes broke existing behavior.
After fixing one light, check the others.
06How should you handle a problem here?
OpenClose
AI starts fixing without a plan
- Where should you look to understand the problem?
- Check whether changes map to feedback IDs.
- Follow these steps to resolve the problem
- Pause further changes, report and preserve work already done, then define scope and passing checks. Do not discard changes merely to return to planning.
- Expected recovery
- Repeat the original action and obtain the expected result; keep unexecuted checks marked unverified.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
I am at “Plan the fix and its passing checks”. Expected: Every fix maps to a feedback ID and repeatable passing conditions.. Actual screen/error: [add here]. Inspect existing files first. Give one reversible action at a time, its location, and expected result. Do not recreate the project or delete existing work.07Material for this step (fill or copy)
OpenClose
If you completed the individual prompts above, do not resend this full template. Use it to organize this step’s material when needed; replace bracketed fields, confirm and copy into the project chat.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
First check access to docs/feedback.md, docs/requirements.md. If missing or inaccessible, identify the prerequisite and owner action, then stop rather than invent prior work.
Read feedback, requirements, current version and checks. Reproduce issues and separate defects, new requests and missing evidence. Label uncertain causes. Write docs/repair-plan.md with issue IDs, impact, minimal scope, passing checks, regression checks and recovery. Explain scope or design changes before acting. Plan only; do not modify business code.Journal practice: fill or supplement this step
For this full-stack practice only: use this combined prompt, fill its fields and send it once instead of also sending the generic template. It does not authorize later steps.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
First check access to docs/feedback.md, docs/requirements.md. If missing or inaccessible, identify the prerequisite and owner action, then stop rather than invent prior work.
Read feedback, requirements, current version and checks. Reproduce issues and separate defects, new requests and missing evidence. Label uncertain causes. Write docs/repair-plan.md with issue IDs, impact, minimal scope, passing checks, regression checks and recovery. Explain scope or design changes before acting. Plan only; do not modify business code.
I selected the full-stack journal practice.
Diagnose this issue read-only. From my URL, input, clicks, expected/actual results and error text, separate facts from missing evidence. Check whether the page sent a request, whether it reached the right backend, the response status/error, and the actual database path and matching record. Inspect logs you can access; for manual collection give exact entry and relevant lines, removing secrets. Do not make me guess the technical cause. Propose the smallest plan with issue ID, cause evidence, files, data impact, backup/recovery, original-case retest and working behavior to recheck. With insufficient evidence request one minimal check. Do not fix, reset or reinstall before my confirmation.08Check this before continuing
Every fix maps to a feedback ID and repeatable passing conditions.
09What should you do if the expected result is missing?
OpenClose
Investigate further rather than defaulting to reinstalling or deleting storage.
Compare with an example (illustration, not your result)
F01: restore capacity on cancellation; reproduce and retest; regress booking and repeated cancel; no styling changes.
0303Fix according to plan and hand back for retestingTo checkOpen this stepClose this step
01What do you need, and what should you keep?
Your approved repair plan and the original failing actions.
Issue status matches your result.
Which files should AI read and save in this step?
In your tool’s project chat, ask it to open the files below. If a file is missing, follow the link back to the step that creates it.
Ask AI to read these materials first: docs/repair-plan.md · docs/feedback.md
Ask AI to save the result to: docs/checks.md · docs/feedback.md
Missing these materials? Return to the step that creates them →
02What will this step help you do?
Fix the agreed issue, then repeat the original failing actions yourself.
Understand it, then judge for yourself
What does this concept mean?
Retesting checks the original issue; trying working features catches unintended damage.
How would you decide in a different situation?
If AI checks pass but the same action fails, what fact should be recorded next?
See a reference answer
Record the actual personal failure at the same revision/steps under the same issue ID.
Answer in your own words; if unsure, compare the explanation above and check it against your own project.
03What exactly should you do in this step?
OpenClose
Project chat and preview
Send one action at a time: select and copy its prompt (⌘C on Mac, Ctrl+C on Windows), switch to your tool’s project chat, paste it, replace bracketed fields and send. Wait for the reply and check the stated result before continuing.
After reviewing the plan, fix one agreed issue.OpenClose
Suggested prompt:
Follow the approved docs/repair-plan.md for 【issue ID】. Preserve data, change only necessary files and report checks plus my retest steps.
Afterwards, you should see: The tool explains the change and retest.
Repeat the original failing actions, then check previously working features.OpenClose
Afterwards, you should see: The issue is fixed and normal behavior still works.
Save your personal retest results.OpenClose
Suggested prompt:
My retest result: 【fill in】. Update docs/checks.md and docs/feedback.md; close only personally verified issues and retain untested items.
Afterwards, you should see: Issue status matches your result.
Practice: personal journal (frontend + backend + database)
Use the same learning-journal project throughout this example. For your own project, compare applicable requirements without creating another project.
Apply one agreed fix, repeat the original failing action personally and check old records remain.
- What should you see after this step?
- The failure, normal create/read and existing data all pass.
- What should you do first if you get stuck?
- If everything was marked passed, restore untested statuses and separate AI checks from personal trials.
04Not sure how? Follow each action
OpenClose
After reviewing the plan, fix one agreed issue.
- What should you see after this action?
- The tool explains the change and retest.
- What should you do if that result is missing?
- Return to planning if the scope expands or a reset is proposed.
Repeat the original failing actions, then check previously working features.
- What should you see after this action?
- The issue is fixed and normal behavior still works.
- What should you do if that result is missing?
- Keep the same issue ID and add evidence if it still fails.
Save your personal retest results.
- What should you see after this action?
- Issue status matches your result.
- What should you do if that result is missing?
- Investigate any difference from the tool’s report.
05Words in this step
OpenClose
Regression testing ↗
Checks whether changes broke existing behavior.
After fixing one light, check the others.
Rollback ↗
Restores earlier software or configuration.
Restoring old equipment does not restore changed records.
06How should you handle a problem here?
OpenClose
The same action still fails after a claimed fix
- Where should you look to understand the problem?
- Verify the new version is open and retain the failed sequence.
- Follow these steps to resolve the problem
- Choose “Retest failed”, add version and actual results, and ask AI to reproduce and update the plan rather than recreate the project.
- Expected recovery
- Repeat the original action and obtain the expected result; keep unexecuted checks marked unverified.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
I am at “Fix according to plan and hand back for retesting”. Expected: AI checks and personal retests are recorded separately; failures return to feedback.. Actual screen/error: [add here]. Inspect existing files first. Give one reversible action at a time, its location, and expected result. Do not recreate the project or delete existing work.07Material for this step (fill or copy)
OpenClose
If you completed the individual prompts above, do not resend this full template. Use it to organize this step’s material when needed; replace bracketed fields, confirm and copy into the project chat.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
First check access to docs/repair-plan.md, docs/feedback.md. If missing or inaccessible, identify the prerequisite and owner action, then stop rather than invent prior work.
Read rules and docs/repair-plan.md. Fix only approved issues and preserve existing work. Rerun failures and related normal flows. Record IDs, version, expected/actual results and unexecuted checks in docs/checks.md. Provide a real preview and personal retest steps. Never mark personal tests passed on my behalf. Save a traceable version; do not publish.Journal practice: fill or supplement this step
For this full-stack practice only: use this combined prompt, fill its fields and send it once instead of also sending the generic template. It does not authorize later steps.
Edit the template with actual material, confirm, then copy into the current project conversation. Mark unknowns as undecided and replace examples as needed.
Confirmed · personalized content ready to copy
First check access to docs/repair-plan.md, docs/feedback.md. If missing or inaccessible, identify the prerequisite and owner action, then stop rather than invent prior work.
Read rules and docs/repair-plan.md. Fix only approved issues and preserve existing work. Rerun failures and related normal flows. Record IDs, version, expected/actual results and unexecuted checks in docs/checks.md. Provide a real preview and personal retest steps. Never mark personal tests passed on my behalf. Save a traceable version; do not publish.
I selected the full-stack journal practice.
Fix only agreed issues; test database changes on a copy first with recovery instructions. Record the same failing case before/after plus old-record reads, without claiming user acceptance.08Check this before continuing
AI checks and personal retests are recorded separately; failures return to feedback.
09What should you do if the expected result is missing?
OpenClose
Return to planning if the scope expands or a reset is proposed.
Compare with an example (illustration, not your result)
F01 / revision def / original case passed / booking regression passed / personally retested → close F01.
Record this check and project version
This is self-reported, not automatic inspection. Old checks do not prove a changed version; complete or report failed and untested work.
Words in this milestoneIn order of appearance; open for the full explanation
References for this milestone
Open these when the question arises; they are not extra required actions.
- Is delivery ready, and in what form?
After acceptance, before real users receive the project.
My project material and backups
Material stays in this browser, not in project files or AI chat. Chinese and English template drafts are separate; project selection and progress are shared. Do not enter passwords or API keys here.
Includes idea, progress, template drafts, personal checks and selection conditions, but not project code. Restore creates a new draft and preserves existing projects.
Download/import unavailable? Restore backup text
Draft and progress stay in this browser.
Complete the actions in order; expand the matching checks and remedies if you get stuck.