issue_reportinglisted

# issue_reporting — the backwards flow An engine defect fixed upstream reaches every fork via `./sc update`; worked around silently, every fork re-derives the workaround. File the issue while the failure is on screen — NEVER batch to session end. A workaround IS a report: deviating from a skill's steps, wrapping a command, or hand-patching state to proceed -> you hold the exact repro; file it now. ## Boundary — engine vs fork | Where | What | |---|---| | **Upstream — file it** | anything the engine materializes/owns: `.super-coder/`, `sc` + every subcommand, engine skills (this catalogue), the boot doc render, the sandbox / dev kit, `./sc update` + migrations, the `_sc` API + `sc mem` | | **Fork — don't** | the repo's app code, fork-local skills (see `local_skill_management`), operator-owned host config | Unsure -> "would the same problem hit any other fork?" yes = upstream. ## Triggers Each row = a real engine defect filed by a fork shell doing ordinary work. Match the left column -> file. | You hit | Real case | |---|---| | A `./sc` command fails out of the box | `./sc verify` always aborted — its own render step needed `SC_ADMIN` it never set (#227) | | A command exits green without doing the work | `./sc test` silently fell back to unittest when pytest was missing — green-washed suites (#219) | | The documented remedy is a closed loop | `./sc lint` said "run `./sc deps` first," but deps skips pip in the sandbox — tool unobtainable from inside the box (#246) | | A