Keep the specification next to its evidence
Living Scenarios shows behavior specifications and their latest test results on Jira work items. Product collaborators can draft scenarios in Jira while the committed feature files remain authoritative in Git. A CLI sends feature files and test reports from CI, so the app does not need repository credentials.
The difficult part is matching changing scenarios to the correct results and keeping derived issue summaries consistent. The implementation handles feature identity, report formats, repeated uploads, asynchronous summary updates, and Jira search properties. A green status is meaningful only when it refers to the right scenario and run.
The real platform changed the implementation
A live Forge staging benchmark ingested 500 feature files containing 5,000 scenarios in 19.9 seconds and uploaded 20,000 results in 18.1 seconds. The workload used generated data and the real CLI, web trigger, and Forge SQL path.
Real deployment exposed behavior the local TiDB environment accepted: large derived-table updates failed inside Forge SQL, so the implementation bounded their batch size. Jira's asynchronous bulk-property writes also required completion polling rather than immediately sending another write. These findings are recorded alongside the measured results.
Release state
The app has been exercised on development and staging sites. The public marketplace release, production operating setup, and some cross-persona and browser validation remain pending. It is a product engineering case study with live integration evidence, not a claim of a launched customer business.
Limitations
- The performance workload is generated test data on a development-tier site, not customer traffic.
- End-to-end in-Jira panel latency was not measured by the ingestion benchmark.
- Marketplace distribution and public CLI publication remain pending.
Source code & evidence
Results reflect the linked evidence snapshots. Source artifacts retain their dates and experimental conditions.
