A diagnosis workflow with the console
This page is for developers and administrators who want a coding agent to find the cause of a problem without touching the installation. It walks through a realistic case with read-only console commands that ship with the kernel. The agent ends with a report; a person decides what to change.
Contents
The setup
- A development or staging copy of the installation, with the problem reproduced.
- A coding agent in a terminal on that copy, told to use only the commands below and the documentation (Coding agents on an Exponential project).
- No production data in the copy, or at least none the agent's model provider may not see.
The case
A site manager reports two things: a setting in an extension's override "has no effect", and an article gives anonymous visitors an "access denied" page.
Step by step
| Step | The agent asks | Command | Why |
|---|---|---|---|
| 1 | What does the documentation say about the setting? | ./console exp:help search "<setting name>" | Know the intended behavior first (Agent-friendly documentation) |
| 2 | Which files set it, in load order, and which value wins? | ./console exp:ini where <file>/<Block>/<Variable> | Shows every file that sets the variable and the value in effect |
| 3 | What is the value now? | ./console exp:ini get <file>/<Block>/<Variable> | Confirms the effective value |
| 4 | May the anonymous user read the article, and which limitation refuses it? | ./console exp:access:check --user=anonymous --node=<id> | Ends with ALLOWED or DENIED and names the policy and limitation |
| 5 | Did the logs record anything around the time of the report? | ./console exp:logs --tail=var/log/error.log --lines=50 | The last entries, with personal data masked by default |
| 6 | Did anyone change the role or the settings recently? | ./console exp:audit search --name=<event name> and ./console exp:audit tail --channel=access | Role changes and setting writes are in the audit trail |
Habits that make the steps reliable:
- One question per step; let the agent read each answer before the next command.
- Use
--jsonwhere a command offers it (exp:access:check,exp:help), so the agent parses rather than guesses. - Add
--allow-root-useronly where the copy runs commands as root, as the guides show.
The report
1. Setting: <file>/<Block>/<Variable> is set in three files. The extension's override sets it,
but settings/override/<file>.append.php sets it again later in load order, so that value wins.
2. Access: the article is in section <n>. The Anonymous role's content/read is limited to
section <m>. exp:access:check names the Section limitation as the refusal.
3. Suggested changes (not made): remove or change the later override; either move the article to
section <m> or extend the Anonymous policy, after deciding which is intended.
4. Pages used: <help page ids>.A person makes the change in the administration or the settings and clears the caches, then asks the agent to run steps 2 to 4 again to confirm.
Why this workflow is safe
- Every command used is read-only:
exp:ini whereandget,exp:access:check("nobody signs in and nothing is changed"),exp:logsreading,exp:auditreading,exp:help. - The agent works on a copy, not production.
- The report names its evidence; a person can run the same commands to check every claim.
Commands that change things (exp:ini set, cache clears, scripts that write) stay with the person.
Related pages
- When something goes wrong
- Finding an error by its reference
- The log viewer
- Free documentation: this page is free to copy, translate and improve under the GNU FDL 1.3.
Become an Exponential Partner
Partners share their information, provide funding, contribute back and develop the products with us, and that is how Exponential grows. A partner website and dashboard are planned for mid-2027: please become a partner today to speed up our growth rate and land new features faster.