AI-Assisted Workflow Extensions#
This draft describes optional AI-supported requirements inspection, not an approved process change or tool qualification. The executable demonstration is described in Execute the Requirements Inspection Demonstrator.
The workflow augments wf__monitor_verify_requirements without replacing the
existing manual inspection. Proposed tool evaluation and monitoring are described
in AI Tool Management.
AI-Assisted Safety Requirements Inspection
|
status: draft
|
||||
An engineer selects feature or component requirements, context, and the applicable public S-CORE inspection checklist. Before sharing input with an AI service, the engineer verifies that sharing is permitted. Roles match Monitor/Verify Requirements, which this workflow augments: the committer remains responsible and approving, with the safety manager supporting as moderator, consistent with Work product Inspections Co.... This also applies to re-inspection triggered by change or release review; no separate reusable sub-workflow is defined for that case. The AI fills the checklist content directly in the exact column structure of
Requirements Inspection Che... (Review ID, Acceptance Criteria, Guidance,
Passed, Remarks, Issue link), parsed from that template rather than
paraphrased, without changing requirements or inventing checklist IDs. Each
entry has a Scope: The existing Proposed review guardrail: After presenting a validated draft, the agent
ends its turn. The reviewer independently checks the assessment, evidence
gaps, parent traceability, and consistency, and may edit the checklist or
request corrections. After changes, the engineer runs the check command
or asks the agent to run it. The prompt instructs the agent to present the
checked result and end its turn. There is no local chat-confirmation command:
the engineer copies the checked checklist into the project’s real
inspection document (for example Proposed constraints:
|
|||||