fix-looplisted
Install: claude install-skill stacktale/stacktale
# Working an error down to zero with stacktale
stacktale writes `errors-ai.log` next to the app's normal logs. Each entry is a complete
report: the root cause first, the app's own culprit frame marked `← YOUR CODE`, the log
lines that led up to the failure, the values involved, and the environment. It is written
for you to read — you should not need to ask the user for a stack trace.
## The loop
`errors_since_last_check` is the whole point. It keeps a cursor per session, so:
1. **Take a baseline.** Call `errors_since_last_check` once. The first call reports what is
currently on file — that is your starting position, not new breakage.
2. **Read the top error properly.** The headline is the *root* cause, not the outermost
wrapper. Go to the frame marked `← YOUR CODE`; that is the app's own code, and it is
almost always where the fix belongs. Read the `story` before deciding — it usually
contains the line that explains *why* the value was wrong, which the stack trace alone
never shows.
3. **Fix it.**
4. **Re-run** the app, or its tests if `stacktale-junit` is on the test classpath (a failing
test writes a report too).
5. **Call `errors_since_last_check` again.** It reports 🆕 for errors that are new since
your last call, 🔁 for ones still occurring, and `✓ No new errors` when the run is clean.
6. Repeat from 2 until clean.
Do not re-read the whole file between rounds. The cursor is there so you only look at what
changed.
## The other tools
- `list_err