ragelisted

# RAGE Research-to-merge loop for one eligible issue at a time. Evidence outranks the inherited design. Load and follow the canonical `adadr` skill for architecture reasoning; do not replace it with a prompt-local approximation. ## Queue 1. Read repository instructions, roadmap/order issue, relevant epic, and open dependencies. 2. Select the first eligible issue by declared priority/order; blockers beat entertaining later work. 3. Do not create a parallel TODO when an issue already owns the work. 4. Keep one RAGE iteration scoped to one consumed issue unless repository policy says otherwise. ## Start evidence Before research or implementation, enter the intended `rage/<issue>-<slug>` branch and initialize the run log with this skill's `scripts/init-run.py <issue> <slug> --repo <checkout>` helper. It records the actual full HEAD SHA and branch, refuses branch mismatches, and refuses to overwrite an existing run log unless an append is explicitly requested. The resulting evidence is: ```org * Run identity - Consumed issue :: #NNN - Branch :: =rage/<issue>-<slug>= - RAGE start commit :: =FULL_40_CHARACTER_SHA= ``` The start SHA is immutable. Keep failed iterations and decisive evidence visible. ## ADADR architecture loop Load `adadr` after queue selection/start evidence and before architecture or implementation work. First execute ADADR's control-plane discovery gate. If the project has no defined/used research-design approval location or workflow, stop this issue and