Local Codex Pet Troubleshooting
See the symptom, surface, and observed state currently held in the browser.
Pet problems look similar on screen even when the cause is a hidden setting, invalid input, unsupported surface, missing skill, or environment mismatch. Describe the visible symptom and route to the smallest relevant check instead of trying unrelated fixes. Start with the local, reversible action described by this practical local codex pet troubleshooting; its evidence boundary remains visible.
LOCAL TOOLUse the first control. No sample output is presented as evidence.
Select the visible symptom and surface to receive one focused, reversible first check. The triage result is a decision aid, not remote diagnosis. It does not inspect the machine, read files, or claim that a product bug has been reproduced.
The workspace reports the selection, current local result, source category, and remaining limitation without presenting an empty field as proof.
See the symptom, surface, and observed state currently held in the browser.
Read one conclusion from the visible inputs.
See which record or test remains missing.


This tool handles one page decision, keeps processing local, and exposes evidence limits, next links, and corrections.
Choose what is visible before choosing a fix.
Begin with a reversible settings, input, or environment check.
Separate documented behavior from community reports and untested assumptions.
Continue into the validator, install selector, compatibility matrix, or focused help page.
The triage result is a decision aid, not remote diagnosis. It does not inspect the machine, read files, or claim that a product bug has been reproduced. Before anyone acts, keep that boundary in plain view.



Check the product surface, Pets setting, selected pet, and current state before changing files.
Use local parsing to isolate syntax or object-root errors before applying profile-specific advice.
Record the product version and whether the session is native Windows or a separate subsystem.
Keep Linux and Windows path contexts separate and name the UI surface that should display the pet.
Confirm the current official creation workflow and skill availability before inventing a manual replacement.
Describe the visible symptom and route to the smallest relevant check instead of trying unrelated fixes. Keep local facts separate from cited guidance, named profiles, dated observations, and unreproduced cause.
Triage starts from what the person can see, not from an assumed package defect hidden behind the symptom.
The router separates visibility, invalid JSON, Windows, WSL, and hatch-pet reports because each needs a different first check.
A version marked Unknown is more useful than a guessed release number that sends the investigation down the wrong branch.
Screenshots can document appearance, but text notes are still needed for product, environment, timing, and steps already attempted.
Preserving the original file and setting state keeps later comparisons possible when a proposed fix makes no difference.
A symptom that follows one account or machine should not be generalized to every user without another reproducible observation.
Activity labels describe what Codex is doing; they do not double as parser severity or package-validation outcomes.
Repeat the same check with the same visible input and compare the results.
Cite the first-party source and date behind a troubleshooting branch.
Name the non-official profile, version, and publisher.
Active work; not verification of troubleshooting branch.
A request for attention; not evidence about unreproduced cause.
An activity state; not installation or rights confirmation.
Interrupted diagnostic route; not resource rejection.
Desktop, Web, CLI, and IDE create different evidence needs for this diagnostic route. Product facts below were checked on 2026-07-26.
Name the intended product and environment before relying on a result.
| Context | Evidence and limitation |
|---|---|
| Desktop | Begin Desktop triage with the Pets setting, wake-or-tuck control, selected companion, activity state, and product version. |
| Web | For web reports, distinguish an official custom image failure from an expectation that a local Desktop resource would synchronize. |
| CLI and terminal | Terminal symptoms require the emulator, graphics protocol, multiplexer state, shell, and rendering observation in the support note. |
| Windows, WSL, and IDE | A Windows host, WSL distribution, and IDE panel are separate contexts; record exactly where the symptom appears. |
This local diagnosis control creates a troubleshooting branch without changing browser history or resolving the unreproduced cause.
Supply only the needed symptom, surface, and observed state.
Create the troubleshooting branch in this tab.
Compare the result with its stated limitation.
Open the matching evidence route.
Clear local input without changing history.
Preserve the original input and current environment before changing a file, setting, or install state. Make one observable change at a time and record the result before continuing.
Keep diagnosis, implementation, verification, and correction as separate stages.
Describe the first observable failure without guessing its cause.
Record product, version, operating system, terminal, and recent changes.
Use the narrowest deterministic check and preserve the original input.
Keep Pass, Warning, Fail, and Unknown separate, then follow the matching guide.
Useful evidence names the claim, resource, version, method, date, result, and limitation.
| Context | Evidence and limitation |
|---|---|
| Symptom captured | The report states what the user saw and when. |
| Surface captured | Desktop, Web, CLI, or IDE is named. |
| Version captured | The product version is known or explicitly Unknown. |
| Input preserved | Original files have not been overwritten during diagnosis. |
| Recent change captured | Updates, setting changes, or new packages are recorded. |
| Next test bounded | The next step can be reversed and has an observable result. |
| Source category | Label official, community, test, report, or Unknown. |
| Checked date | Date time-sensitive evidence. |
| Version | Record affected product and resource versions. |
| Limitation | Name what remains unproven. |
| Correction route | Keep a path for evidence challenges. |
| Privacy review | Remove secrets, personal data, and private paths. |
Official facts, community profiles, tests, and unknown claims stay separate.
This evidence track cites the strongest source and leaves missing facts visible.
Use current first-party Pets guidance for product behavior and workflow facts.
Read current sourceName the non-official format source, version, and checked date.
codex pet formatRecord a reproducible environment, method, outcome, and limitation.
codex pet compatibilityRoute missing, stale, or contested evidence to verification and contact.
contact codex petsThese visible answers describe current behavior and limits; no FAQ structured data or search appearance is promised.
Start with the visible symptom, selected surface, and current official behavior. Avoid editing package files until a deterministic input problem is found.
No. Selections stay in local React state and only choose a relevant checklist.
Usually not. Confirm settings, surface, input validity, and environment before introducing a second change.
Include the product, version, environment, steps, expected result, observed result, and any non-sensitive validator report.
No. Your input stays in local browser state while this page creates the troubleshooting branch.
No. This independent diagnostic route cites first-party documentation only for the specific product facts it describes.
Unknown means the available evidence is not sufficient to support a conclusion for the current context.
Recheck the troubleshooting branch after its product, environment, source, permission, or review method changes.
Every link reaches a real P0 route with descriptive anchor text and a matching intent.
Return to the first control, add only the symptom, surface, and observed state, and compare the troubleshooting branch with its limit.