Dashboardopencode-loreEvidence-driven repository, testing, an…

Evidence-driven repository, testing, and review workflow

Category: preference
Confidence: 0.80
ID: 01a0aeaf-54ae-7ea7-a34d-c617bdb1126b
Project ID: 6f4be9ff-ed84-4cca-a9e7-732a0b0b8677
Cross-project: No
Recalled in other projects: 0
Source session: 1YzADfrYkhbUW0CqL
Created: 2026-09-17 09:23:15
Updated: 2026-09-17 09:25:21

Content

Use the established branch/PR workflow and review code before committing. For every bug fix, add a deterministic regression that exercises the real production precondition, prove it fails on the base, passes with the fix, and fails when the guard is removed. For nontrivial changes, perform separate correctness and security reviews with exact current file:line evidence. Freeze the candidate identity and hashes before exhaustive validation; review it read-only without network, services, formatting, or VCS mutation. Run focused checks, the exact aggregate gate, fresh bundle verification, and required lint/typecheck/format checks with recorded exit codes; shard only if the aggregate result is inconclusive. In .jj repositories use noninteractive jj mutations, verify jj st/log, and treat conflicts as recorded state. Resolve findings in a new child and refreeze.

Move to: