cambridge-beer-festival-app: Skill for Claude Code

Diagnostics and Tooling

Every fact below was re-verified against the working tree on 2026-07-02 (see Provenance). If a command's output disagrees with what's written here, trust the command — then fix this file.

Theme: run the measurement tool once, read its output carefully, and interpret it correctly. Don't guess at "it's probably fine" or "that looks like the bug" — grep the log, decode the stack, read the coverage line.

---

1. Test / analyze log workflow

./bin/mise run test [path] and ./bin/mise run analyze [path] are thin wrappers (mise-tasks/test.sh, mise-tasks/analyze.sh) around flutter test / flutter analyze. Both:

1. Print a log path FIRST, before running anything: TEST_LOG=/tmp/test-XXXXXX.log or ANALYZE_LOG=/tmp/analyze-XXXXXX.log (mktemp-generated).
2. Stream output through tee into that file, filtering two noise lines (Woah! You appear|superuser privileges — root-user warnings irrelevant here).
3. Preserve the real exit code via ${PIPESTATUS[0]} (the tee would otherwise mask a non-zero flutter exit).
4. Print a ready-to-use grep command at the end:
   • test: grep -n 'FAILED\|ERROR' $TEST_LOG
   • analyze: grep -n 'error\|warning' $ANALYZE_LOG

Run once, grep many times. Do not re-run the whole suite to look for a different string — the log file already has everything.

./bin/mise run test test/domain/services/drink_filter_service_test.dart
# TEST_LOG=/tmp/test-AbCdEf.log
grep -n 'FAILED\|ERROR' /tmp/test-AbCdEf.log
grep -n 'preference_keys' /tmp/test-AbCdEf.log   # any other query, same log

Stable paths across multiple runs

Override the mktemp path with the env var so repeated runs land in one place you can diff:

TEST_LOG=/tmp/my-run.log ./bin/mise run test
ANALYZE_LOG=/tmp/my-analyze.log ./bin/mise run analyze lib/screens/

What good/bad output looks like

• test: flutter test prints one line per test group in MM:SS +N: <name> form, ending All tests passed! (good) or Some tests failed. with FAILED markers and stack traces above the summary (grep target: FAILED\|ERROR). Verified live: 00:02 +2: All tests passed!.
• analyze: Analyzing <paths>... then either No issues found! (ran in N.Ns) (good — verified live against lib/utils/ and lib/constants/), or one line per finding, bullet-separated: info • <message> • <file>:<line>:<col> • <lint_rule_name> (severity is info, warning, or error; since #524 the task runs without --no-fatal-infos, so every severity fails it).