The hypothesis-debugging skill
plans a debugging investigation rather than performing it. It writes a
falsification plan and hands execution to a sub-agent, preferring alchemist.
Plans are written under docs/debugging/, which the skill creates when it is
missing. Each plan is named
debugging-plan-<year>-<month>-<day>-<problem-slug>.md: a four-digit year,
zero-padded month and day, then a lower-case, hyphen-separated slug naming the
problem under investigation. For example, a plan opened on 20 August 2026 for a
faulty ACP skill agent menu is
debugging-plan-2026-08-20-acp-skill-agent-menu.md.
The slug is required. The date alone sorts plans chronologically but leaves a
directory of indistinguishable filenames, so the slug is what makes a plan
identifiable without opening it. Earlier plans used an opaque
debugging-plan-{timestamp}.md name; see the
migration guide for renaming them.