pflow-golang-troubleshoot

If troubleshoot-context.sh fails (non-zero exit or "status":"error" in its JSON) print ⚠️ <error.message> and stop. Other failing commands are evidence — record and continue.

Steps

1. Run .agents/skills/pflow-golang-troubleshoot/scripts/troubleshoot-context.sh → {toolchain, goos, goarch, cgo_enabled, debug_tools[], task_runner, test_targets[], test_files, test_dirs[], test_build_tags, hint}.

2. Read references/methodology.md, then the file for the symptom:

   Symptom	Read
   panic, wrong output, nil pointer, index out of range	common-bugs.md
   hangs, deadlock, race report, leak, "all goroutines are asleep"	concurrency-debug.md
   slow, high memory, high CPU, GC pressure	pprof-delve.md
   test fails, flaky, passes alone but fails together, coverage gaps	testing-debug.md
   does not compile, go mod / vet / lint errors	build-errors.md

1. Reproduce first, with the narrowest command available: a test_targets entry from the project runner, else go test -run 'TestX' -race -count=1 ./pkg/.... Respect test_build_tags (e.g. -tags e2e needs its backend — ask before running).
2. Form one hypothesis at a time, prove it with evidence (log, test, profile, dlv), fix the root cause, add a regression test, rerun the reproduction and the project gate.
3. Report: symptom → root cause → fix → proof (test name / command output). Never report "fixed" without the reproduction passing.

Paths are relative to .agents/skills/pflow-golang-troubleshoot/.

Gotchas

• hint non-null means CGO_ENABLED=0; -race may need CGO_ENABLED=1 go test -race.
• Missing dlv/benchstat/govulncheck in debug_tools: propose go install …@latest or go run …@latest; don't assume they exist.
• Don't change behavior to make a symptom disappear; if the fix widens beyond the cause, say so and stop for confirmation.