Glossary
Look up the key terms used across Guardian.
Definitions of terms you’ll encounter across Guardian.
Structure
| Term | Definition |
|---|---|
| Organization | The top-level account in Guardian. All projects, members, and policies belong to an organization. |
| Project | A distinct scope of work — a product, initiative, system, or agent — that your team wants to bring into compliance. |
| Policy | A structured set of compliance requirements. Policies may represent regulations, RMFs, or internal SOPs. |
| Applied policy | A policy that has been applied to a project, activating the policy’s controls and enabling your team to start tracking compliance. |
| Policy Library | The collection of all policies available to apply to your project. The Policy Library is scoped to your organization. With agent certification on, the sidebar calls it Library, which holds Policies and Programs. |
| Control | A discrete mandate within a policy, or one requirement within a Governance Program. |
Compliance workflow
| Term | Definition |
|---|---|
| Declaration | A statement describing how a control has been implemented, or, in an assessment, how an agent meets a program’s controls. |
| Review | A statement describing whether a control’s implementation is satisfactory, or whether an agent meets a program control. |
| Attachment | A file submitted to support a declaration or review. |
| Control status | A snapshot of a control’s current state. A control’s status updates based on the last action taken on that control. |
| Credential | An attestation that a project has met a specific policy. An individual agent’s counterpart is an agent certification. |
| Report | A point-in-time record of a project’s compliance status toward a specific policy. |
| Integrity action | An action that is cryptographically signed and hashed at the moment you take it, making it permanent. |
Indicators
| Term | Definition |
|---|---|
| Indicator | An automated evaluation that determines whether a system, environment, or function matches machine-readable criteria you define. |
| Event | Data your systems send to an indicator’s ingestion endpoint to be evaluated. |
| Ingestion endpoint | The endpoint an indicator connects to. |
| Extraction Rule | A rule that tells the indicator where to look in an event (using a JSONPath) and what value to expect there (using Success Criteria). |
| JSONPath | An expression used to select a value from incoming event data. |
| Success Criteria | The conditions that determine whether an evaluation results in Success or Failure. |
| Evaluation status | The result of an indicator’s most recent evaluation (or lack thereof): Awaiting Evaluation, Success, Failure, or Evaluation Error. |
| Declaration Template | The template used to define the declaration an indicator submits to the controls it’s applied to when an evaluation results in Failure. |
Agent certification
| Term | Definition |
|---|---|
| Governance Program | An organization’s standard for certifying an individual agent: a set of controls, the configuration fields a certification binds, and how long a certification lasts. |
| Program version | One immutable version of a Governance Program, addressed as slug@version and identified by its Manifest CID. A change to a program is published as a new version. |
| Assessment | An agent’s case for certification under one program version: the pinned configuration and operating profile, declarations, and not-applicable claims. |
| Submission | A numbered, frozen copy of an assessment’s draft, sent to the program’s reviewers. A later submission supersedes it. |
| Agent certification | A signed credential that one agent, running one configuration, met every control of one program version. It never covers a project or a later configuration. |
| Configuration | The fields of an agent’s setup that its requests capture, such as its permissions, skills and settings files. |
| Configuration CID | The fingerprint of a configuration. The same settings always give the same CID, and a change gives a new one. |
| Operating profile | The owner’s versioned declaration of what an agent is for, the systems it reaches, and what it must never do. |
| Configuration standing | How an agent’s keys stand against the configuration a certification names: Current, Partly Drifted, Drifted, or Idle. Guardian shows None or Can’t Tell when there is no standing to read. |
| Drift | An agent instance running a configuration its certification doesn’t name. Wherever a certification is required, the gateway refuses that instance’s requests unless another certification names what it runs. |
| Cordoned | The state of an agent key whose newest request ran a configuration its certification doesn’t name. It clears once the key runs a certified configuration again. |
Integrity and verification
| Term | Definition |
|---|---|
| CID | A CID (Content Identifier) is a unique fingerprint derived from a piece of content. |
| DID | A DID (Decentralized Identifier) is a unique identifier for a user, organization, or even system, used to sign actions such as declarations and reviews. |
| UUID | A UUID (Universally Unique Identifier) is a standard identifier used to distinguish records such as projects and evaluations. |
| Governance Manifest | A machine-readable file (manifest.json), included in a report’s .zip, that lists the CIDs and DIDs of every declaration, review, and attachment so the report can be independently verified. |
| Verified | A record is verified when its cryptographic proof is confirmed intact — its content matches its original CID and its signature traces back to the signer’s DID. |
| Certified | The mark an applied policy carries once it has an active credential: a seal after its name at the top of its Overview, and a badge beside its name in the project’s list of applied policies. For an agent’s key, the state of one whose newest request ran the configuration its certification names. |
| Lineage Manifest | A cryptographically verifiable record of how a dataset, model, or other asset is produced. This file can be viewed as a Lineage Graph in Guardian. |
Related
- Key Concepts — fuller explanations of the core concepts and the actions tied to them
- Integrity Actions — which actions are signed, permanent, and included in a report
- Indicators — configure extraction rules, success criteria, and declaration templates
- Reports — what a report’s .zip and governance manifest contain
- Lineage Graphs — view a lineage manifest as an interactive graph
- Agent Certifications — what a certification covers, and how drift is enforced