Niyatna Docs

Proof of Intent & Verification

Review execution logs, file diffs, credentials, and staging outcomes before committing changes.

The Verification Layer (accessed via Niyatna HQ) is the core security surface of Niyatna OS. It gathers all evidence produced by agent runtimes and packages them into a verifiable format.

Under the Proof of Intent standard, human operators review staged outputs and audit execution logs before any changes are committed to production systems.

Niyatna proof editor

Verification Primitives

The Verification Layer aggregates several proof types:

  • Filesystem Diffs: Line-by-line file modifications staged by the agent sandbox.
  • Database Transaction Diffs: Staged SQL changes showing before-and-after tables.
  • Execution Logs: The raw stdout/stderr logs from terminal commands run by the agent.
  • API Receipts: Payload responses from integrated third-party platforms (HubSpot, Stripe, Slack).
  • Rendered Previews: Screenshots or live viewports of UI components and document templates.

The Rule: Evidence Over Trust

In Niyatna OS, confidence scores and summaries do not equal proof. If a human operator cannot inspect the exact evidence, it is not treated as complete. The Proof of Intent standard guarantees that delegated execution maps back to the founder's intent before any write action is approved.

OpenIntent & Intent Primitive

The primary primitive for translating a human business goal into a structured, verifiable task contract.

Delegation & Proof Trails

Track filesystem modifications, command outcomes, and API changes back to the original intent.

On this page

Verification PrimitivesThe Rule: Evidence Over TrustIn Niyatna OS, confidence scores and summaries do not equal proof. If a human operator cannot inspect the exact evidence, it is not treated as complete. The Proof of Intent standard guarantees that delegated execution maps back to the founder's intent before any write action is approved.