Property-Level Reconstructability of Agent Decisions: An Anchor-Level Pilot Across Vendor SDK Adapter Regimes
For developers and auditors of agentic AI systems, this pilot provides a preliminary measure of how well agent decisions can be reconstructed across different regimes, but it is incremental due to its small scale and single-annotator design.
The paper evaluates the reconstructability of agent decisions across six SDK regimes using a Decision Trace Reconstructor, finding that per-property reconstructability varies from 42.9% to 85.7% and identifying one regime-independent gap (reasoning trace) and four regime-dependent gaps.
Agentic AI failures need post-hoc reconstruction: what the agent did, on whose authority, against which policy, and from what reasoning. Cross-regime feasibility remains unmeasured under one property-level schema. We apply the Decision Trace Reconstructor unmodified to pinned worked-example anchors from six public vendor SDK regimes spanning cloud-agent, observability, tool-use, telemetry, and protocol traces, plus two comparator columns. Each Decision Event Schema (DES) property is classified as fully fillable, partially fillable, structurally unfillable, or opaque. Per-property reconstructability of an agent decision already varies between regimes at this anchor scale. Strict-governance-completeness separates into three tiers ranging from 42.9% to 85.7%, yielding one regime-independent gap (reasoning trace), four regime-dependent gaps, and one Mixed property; the pilot is single-annotator, one anchor per cell, descriptive, with outputs checksum-verifiable from a deposited reproducibility package.