fixlisted

# /mneme:fix — Diagnose a bug into a runnable fix phase The DEBUGGING entry of the lineup: grill interrogates, plan plans, dev executes — fix DIAGNOSES. It takes a bug in plain words («не работает X», «упало в проде», «тест красный после мержа»), reproduces it as a RED minimal test, fans out 2-3 hypotheses with evidence for a digit choice, and — after the user picks a direction and approves the draft — emits a one-phase micro-spec `docs/FIX-<slug>.md` whose done-when gate IS the regression test, migrates it, and ends at a ready `/mneme:dev fix-<slug>` command. The FIXING itself is dev's job, through the engine — with the event log, attempts, resume, and the `resolved_error` / `fixed_test` harvest that the machine already provides. fix is to bugs what plan is to features: the entrance, never the executor. Reference: the reproduce → minimize → hypothesize → fix → regression-test cycle of /diagnosing-bugs (mattpocock/skills). The CYCLE is borrowed; the protocol is ours — digit fans, hard stops, the engine's executable gates instead of honor-system discipline. ## Arguments `/mneme:fix "<bug in words>"` — the observed failure, described in plain language. Paste what you have: an error text, a failing test name, a «работало — сломалось» story. Examples: - `/mneme:fix "npm test падает на check-workflows после переименования шага"` - `/mneme:fix "recall возвращает пусто, хотя заметки есть"` No argument → ask for the bug in one line (numbered prompt not needed — it is the sole o