🤓 Latest Submissions
ShipSafe AI — Continuous Release Safety
ShipSafe AI is a continuous release-safety platform designed to help developers understand the consequences of a software change before it reaches production. Instead of functioning as a generic AI code reviewer, ShipSafe treats a software repository as a connected system and investigates how a change can affect implementation, tests, security, requirements, APIs, databases and dependencies. The workflow begins when a monitored GitHub repository changes through a push or pull request. ShipSafe collects deterministic evidence from the actual repository, including the commit and diff, changed files, project structure, tests, documentation, requirements, API definitions, database schemas, migrations and dependency manifests. This evidence forms the context for specialized analysis agents. Five specialist agents investigate the change in parallel: a Change Impact Agent identifies affected components and workflows; a Test Gap Agent searches for missing regression coverage; a Security Agent investigates security-sensitive changes; a Requirements and API Agent compares implementation against specifications and external contracts; and a Database and Dependency Agent checks persistence, migrations and dependency consistency. A Release Synthesizer then consolidates the verified findings into an auditable release decision. ShipSafe classifies a release as READY, ATTENTION, or BLOCKED based on evidence-backed findings and validation results. Developers can inspect the exact affected files and evidence, view an impact map, trace requirements to implementation and tests, simulate potential regression paths, and receive prioritized remediation recommendations. After a fix is applied, ShipSafe can re-run validation and compare the before-and-after state. A key design principle is evidence over assumption. Agents are not allowed to invent files, tests, requirements or database structures. Findings are tied to the repository and commit being analyzed and are validated against actual source evidence before appearing in the dashboard. The product uses a conventional engineering dashboard rather than a chatbot-style interface. A Python backend handles repository analysis and event orchestration, while Streamlit provides the release monitoring and investigation interface. Grok is used as the live reasoning layer for the specialist agents. IBM Bob 2.0 was used extensively during development to design and build the agentic workflow, orchestrate specialized subagents and parallel analysis, work with project documentation, and develop the release-analysis architecture. ShipSafe's goal is simple: instead of discovering what a change broke after deployment, developers can investigate its potential impact before they ship it.
27 Sep 2026
