explainlisted

# explain — decompress spec into prose Inverse of the `glyph` skill (math-glyph encoder). Human-facing. Reads SPEC.md, expands one citation into plain English with cited context. Writes nothing. ## LOAD 1. Read `SPEC.md`. If missing → "no spec, nothing to explain." Stop. 2. Parse `$ARGUMENTS`: - `§T.n` / `§V.n` / `§B.n` / `§I.<key>` → that row - `§G` / `§C` → that section in full - `--next` or empty → lowest-numbered §T row with status `.` 3. Renumber-map probe — `.claude/spec-renumber-map.json` ∃ (written by reorganize skill per §V renumber permission) → on `§V.<n>` arg, walk `old:V<n> → new:V<m>` chain newest-first until ⊥ further mapping, then resolve substituted id against current SPEC.md. Read-only contract preserved per read-only-diagnostic invariant — map consulted, ⊥ mutated. ⊥ exist → arg resolves directly against current SPEC.md. 4. If citation absent → list valid ids in target section. Stop. ## EXPAND For the chosen citation: 1. Quote the raw math-glyph line(s) verbatim in a code block. 2. Restate in plain English. No math-glyphs, no fragments — full sentences. 3. Pull in cited siblings: - §T row → expand every §V and §I it cites. - §V row → list §T tasks that cite it and §B bugs that reference it. - §B row → expand the §V it broke and the fixing §T. - §I row → name §V invariants that constrain it. - §G / §C → no cross-cites; just prose. 4. Close with one line: what the reader should now understand. ## OUTPUT SHAPE ``` ## §T.<n> — ad