explain-codelisted

# Explain code Explain this repository from what was read in it this session, and say how sure each claim is. The enemy is the fluent answer built from a name, a comment, the asker's own theory or memory, which the user then acts on. The overcorrection is an answer so hedged or so long it decides nothing, or a list of locations where a narrative was asked. ## When to use - `how`: how something works, a walkthrough before a change, where a value really comes from. - `why`: why code is shaped this way, whether a stated reason holds, whether something can go. - `teach`: how plus why, pitched at the asker's level. - The skill changes no file; a change the answer leads to goes to the stage that owns it. - Not for locations only: the `exo:locate-code` agent returns those. - Not for how an external library or API behaves: `check-docs` owns it. - Not for behavior reported wrong with an unproven cause: `find-cause` owns it. - Not for where the architecture should change: `audit-architecture` owns it. ## The loop 1. **Name the mode and the question.** Write the question in one sentence and pick `how`, `why` or `teach`; a theory the asker brings is a hypothesis to test, never a finding. 2. **Read before you claim.** Every claim about behavior rests on a file and line opened this session, because a name, a comment or a recalled pattern can say the opposite of the code. 3. **Trace the whole chain.** Follow the path from entry to effect, including config, environment and call-site ove