Agentic framework for integration, end2end, multiuserinteraction, security testing based on usecases.
Answers the two questions gating the rest of TD-WP-0002. Classifier: stratify evidence into S1 surface / S2 realization / S3 judgment. Adaptation writes only to S1; claims and invariants are run inputs with no adaptation write path, so False Adaptation Rate = 0 is architectural rather than a tuning target. Classification is a total function over three signals, defaulting to AMBIGUOUS on missing evidence; a surface change concurrent with a verdict change is PRODUCT_DEFECT. Semantic actions declare permitted surfaces so surface-substitution cannot pass as recovery. Provenance: claims require a source causally independent of the implementation. agent-from-implementation output is an exploratory hypothesis until a human promotes it. New adoption precondition D-07: test-driver requires an independent observation channel into the system under test. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Assistant: claude-code Assistant-Model: opus Assistant-Process: 1629012@bnt-lap001 Assistant-Session: 78d4fb13-8a1e-474b-87a3-9b9261c49a39 |
||
|---|---|---|
| docs | ||
| history | ||
| research/findings | ||
| workplans | ||
| .custodian-brief.md | ||
| .gitignore | ||
| .repo-classification.yaml | ||
| AGENTS.md | ||
| INTENT.md | ||
| README.md | ||
| SCOPE.md | ||
| WORK-RECORDS.md | ||
test-driver
Agentic framework for integration, end2end, multiuserinteraction, security testing based on usecases.