Data policies and agent context
Connect requirements, datasets, and implementation evidence in one review.
Data policies give your team and its coding agents a shared record of how data should be handled. Add source requirements, review their meaning, and map them to the datasets used by your pipelines.
Governance Data policiesExampleClinical data use
Requirement 01 · Clinical records
Clinical records may be used only while the member has active consent for this purpose.Clinical data policy · Section 2
Use the member’s current consent. Exclude records when consent is missing or withdrawn.
Reviewed- Consent source
- clinical.consent
- Mapped dataset
- clinical.records
- Person identifier
- member_id
How should recovery copies be handled?
- 01Capture
- 02Review
- 03Map data
- 04Use with agents
Policy workflow
Add requirements
Add a policy or internal specification in Governance. Keep each requirement linked to the source excerpt that supports it.
Review and map
Confirm the intended behavior and select the relevant catalog tables. Record unresolved questions before asking an agent to implement the requirement.
Work with your coding agent
Open Agent context and copy the policy reference into your connected agent. Ask it to inspect the pipeline, propose changes, run tests, and report which requirements it checked.
What a policy contains
| Item | What it gives your team |
|---|---|
| Source excerpts | The original requirement behind each interpretation. |
| Reviewed decisions | Agreed behavior and questions that still need an answer. |
| Dataset mappings | The catalog assets the requirement applies to. |
| Implementation evidence | Reported checks tied to a policy, data scope, and code revision. |
A saved policy provides context. Applying it to ingestion, access, or deletion requires an implemented and reviewed workflow.
When requirements change
Requirements stay linked to their source and mapped data. Changes to an approved requirement return it to review. Agent-reported checks remain labeled as reported evidence, rather than independent verification.
An assessment records the policy revision, data scope, and implementation revision it checked. If that context changes, the agent needs to refresh it and reassess. Requirements omitted from an assessment remain unchecked.