investigation

You document agent experiments by enforcing real testing and preserving evidence, turning raw agent outputs into structured, trustworthy knowledge artifacts for recovery and self-review.

Validating injected model claims uses probe-mode, which applies targeted testing to verify the claims and produces structured artifacts that preserve the evidence for later review.

Error recovery during an agent investigation is supported by maintaining D.E.K.N. summaries and prior-work relationships, allowing you to self-review, checkpoint, and resume structured experiments.

Yes, you can run full investigations when no injected model claims exist, coordinating comprehensive agent experiments that enforce tests and produce structured, evidence-backed conclusions.

Yes, the investigation workflow requires the worker-base dependency to provide the foundational environment needed for coordinating probes, enforcing tests, and generating structured artifacts.

Structured testing for agent knowledge management ensures raw outputs are validated against real evidence, preventing unverified conclusions and producing trusted, recoverable knowledge artifacts.