From our inbox

A follow-up scout tested whether report claims stayed current while the inbox loaded another page. It held the page response, claimed an existing report, and then released the response. The new claim disappeared because the loader restored an older copy of the list.

A skills-store scout found an alert skill that kept its inventory, setup instructions, and operating guide in one long body. It had no bundled reference files. Agents had to load the full body before they could choose the information their task needed.

Error tracking caught a crash when an inbox report tried to show a pull request diff. The file summary called a string method before it checked whether the response contained a diff. That exception replaced the whole report page.

A code quality scout found that the IQR alert detector treated a zero multiplier as a missing value. The fallback replaced zero with the default and widened the alert boundaries. A small test case showed that this could hide an anomaly that the chosen setting should detect.

A scout that watches the CI merge queue repeatedly starts with a query that cannot run. Its instructions compare a timestamp stored as text with a date value. The scout finds a working conversion during its runs, but the saved query still contains the same mistake.

A custom scout found that DateTimePicker stories build date ranges, labels, and calendar bounds from the real clock. Screenshots therefore change with the date even when the component code stays the same. These differences add noise to visual reviews.

The Twilio source guide still warns that the source is in alpha. A custom scout compared the guide with merged product code. The code marks the source as generally available, but the old warning remains in the documentation.

A saved chart stops before the end of its date range. A dashboard scout runs the same query with a larger row limit and finds the missing recent dates. The query returns older dates first, so its default row limit excludes the newest results.

A daily digest scout saves new memory entries under each day's date. Those entries repeat work boundaries that its existing progress markers already retain. A scout review traced the growing memory to instructions that require new dated entries on every run.

A custom scout found repeated requests to /feed.xml that returned 404. It checked the live URLs and found that /rss.xml still served the feed. Other legacy feed paths already redirected there, but /feed.xml had no redirect rule.

A GitHub issue showed that opening the inbox creates an automatic maintenance scout. The setup checks treated that scout as proof that a person had enabled self-driving. Projects with only automatic scouts skipped the introduction and later lost their background scout.

Invalid JSON in ~/.claude/settings.json stops the PostHog Claude Code plugin from installing. The wizard shows the command's error text without explaining what to fix. Error tracking and a code review also show that changing command text creates separate issues for the same failure.

A warehouse scout compares run records with completion events to check for missing data. Its instructions say to compare complete hours, but the query starts and ends partway through an hour. The two sources use different timestamps, so those partial hours create a false gap. The scout traced the discrepancy to its own comparison query.

A GitHub issue led the investigation to a mismatch between warehouse syncs and signal collection. Each sync gives every open pganalyze issue a new sync timestamp. The signal collector treats that timestamp as evidence of a new issue, so unchanged issues enter processing again.

The Android configuration table describes maxRetries as the retry limit for failed requests. A custom scout compared that description with the released SDK source. The setting only limits retries for push-subscription registration failures. It does not control event, session recording, or log ingestion.

A GitHub issue reported that the Scouts search box removes a trailing space when typing pauses. The investigation traced this to the search text being trimmed for the URL, then copied back into the input. Typing "checkout ", pausing, and typing "errors" therefore produces "checkouterrors".

The docs scout finds pages that still call a released feature alpha. It verifies the release in product code and identifies the exact text to change. Its instructions still classify these mechanical corrections as needing human input, so a ready-to-write fix waits for a decision that the evidence already answers.

The dashboards scout compared the Signals pipeline dashboard against the data it could plot. The dashboard already covers funnel outcomes, dismissal feedback, and safety blocks, but no tile reads the view that records signals dropped before they reach a report. A drop in one source could grow for days before anyone noticed.

pganalyze flags a slow read on the activity log table. The notification feed filters a team's whole history by a very selective set of scope and item pairs, then takes the 10 newest rows, so the ordered walk reads far down the history before it fills one page. The same request also fires several near-identical unbounded reads over the table. Nobody sees an error, just a slow bell, and the cost grows with every team's history.

While checking its own coverage, the logs scout found that almost all PostHog AI log lines come from a background worker service, not the web service its instructions tell it to read. Its main query only reads the web service, so it cannot see new log patterns or rate changes where most of the traffic actually lands. The web service still writes some scoped error lines, so it stays useful too.

In the desktop app, WorkColumn calls selectedTaskIds.includes(item.id) inside the loop that renders each row. The selection is a plain array, so every lookup is linear, and a long list with many selected items does quadratic work on each render. The selection hook already builds a Set for filtering, so the fix has a pattern to follow.

A burst of users opening the similar issues panel got a 500 instead of a 400. The similar issues query runner passes the issue id straight into a Django ORM filter, and the query API only maps DRF validation errors to a 400. It is the only error tracking runner that skips the shared validate_uuid_param guard its four siblings call. A second, quieter problem sits in the same runner: a correct 404 for an issue with no fingerprints counts against the query SLO and gets captured as an error.

A GitHub issue reported that adding a new line to an Inbox refund note submits the form. Once a reason is selected, both Enter and Shift+Enter reach the form's submit handler. The investigation found that the note field lacks the keyboard event guard already used by the dismiss dialog.