Skill Claude CodeCodex
Add a test for LLM/agent behavior in the right layer - use when adding an eval for a model-driven behavior, when unsure whether something needs a judge or an assert, or when triaging an eval that has gone red.
Skill Claude CodeCodex
Run timed spoken reps where the user defends their own technical decisions under hostile questioning, with a collaborative walkthrough mode for what they can't yet answer - use for rehearsing design reviews, talk Q&A, demo defenses, or any setting where engineering judgment must survive out loud.
Skill Claude CodeCodex
Graduate a procedural note from the user's vault into an agent-executable skill - use when a note's steps keep being followed by hand, when the user says a procedure "should be a skill," or when auditing the vault for graduation candidates.
Skill Claude CodeCodex
Distill an insight or story from the current conversation into a note in the user's Obsidian vault - use when the user says "save this", "capture this", or "add this to my notes", or when another skill needs to record into the vault.
Skill Claude CodeCodex
Research how an industry, market, or domain actually does something and benchmark our situation against it - use when the user names a context (e.g. "Taiwan ecommerce", "testing principles") and wants common practice researched, asks how our approach compares to the norm, or whether a divergence from standard is a…
Skill Claude CodeCodex
Translate the technical discussion at hand into what a user will actually feel - use when the user asks what a change means for users, wants the UX framing of a technical decision, or needs to explain engineering work to a non-technical audience.
Skill Claude CodeCodex
Decide whether a decision deserves an ADR and write it in a form that survives re-litigation - use when a real trade-off was just settled, when asked "should this be an ADR?", or when recording an architectural decision.