phygitai_audit
ChatGPTUse to review one user-described AI-assisted product project across G0–G5. Only evaluate supplied records; do not verify tests or certify safety.
Overview
A research-stage, public screening tool for designers, students and researchers examining AI-assisted product design. Provide project descriptions and attributed evidence; use six modes (Audit, Compare, Inspect, Detect, Explain, IBR) to organize records, trace human decision authority, identify unsupported claims and document provisional discrepancies. Uses a separate read-only MCP server to process the evidence submitted for each request. No independent device monitoring, empirical validation, safety testing, certification, or scores. The server does not access private research methods or the original private PhygitAi backend.
Tools
phygitai_build_intention_behavior_record
ChatGPTBuild a structured seven-field Intention–Behavior Record (IBR) for one design decision, using user-supplied field statuses and evidence source IDs. No human approval or independent verification is implied.
phygitai_compare
ChatGPTUse for qualitative side-by-side evidence coverage in two separate projects. No scores, rankings or winner.
phygitai_detect
ChatGPTClassify explicitly provided intention-versus-outcome discrepancies only provisionally. Missing observations are not a fault and not Category C.
phygitai_explain
ChatGPTExplain an existing finding and its user-supplied source labels without changing classifications or running a new audit.
phygitai_inspect
ChatGPTIdentify which user-supplied claims and source labels are present, planned or missing. Does not inspect external systems.
Example Prompts
Click any prompt to copy it.
