Behavioral classifications
Behavioral classifications describe what an interpreter does and how it should be understood by users when composing Courier jobs. These classifications focus on intent, effect, and risk, independent of how the interpreter is implemented.
These categories are descriptive, not prescriptive. An interpreter may fit cleanly into one classification or exhibit characteristics of more than one. In all cases, the interpreter contract must make its intent and guarantees explicit.
Assessment and evaluation interpreters
Assessment interpreters evaluate state without making changes.
These interpreters are used to answer questions such as:
- Is the system compliant?
- Does the current state match policy?
- What risks or issues are present?
Typical characteristics:
- Read-only or observational behavior
- Deterministic results given the same inputs
- Rich structured output and evidence artifacts
- Common use in reporting, governance, and audit workflows
Examples of outcomes include compliance posture summaries, drift detection reports, and readiness or validation results.
Assessment interpreters often serve as decision points for downstream remediation or approval processes.
Remediation and action interpreters
Remediation interpreters perform controlled changes to achieve a desired outcome.
These interpreters are used to:
- Correct detected issues
- Apply approved changes
- Bring systems into a known or compliant state
Typical characteristics:
- Explicit, constrained actions
- Clear preconditions and post-conditions
- Structured reporting of what was changed
- Strong emphasis on safety, traceability, and the ability to audit
Remediation interpreters must clearly signal when changes occur and produce output suitable for change records and governance processes.
Discovery and collection interpreters
Discovery interpreters collect data needed for evaluation, reporting, or diagnostics.
These interpreters are used to:
- Gather facts, metadata, or diagnostics
- Inventory configuration or system state
- Produce evidence bundles for downstream analysis
Typical characteristics:
- Data-centric structured output
- Emphasis on artifact production
- Minimal interpretation or transformation
- Designed for reuse across multiple workflows
Discovery interpreters frequently precede assessment or remediation steps.
Orchestrated or multi-phase interpreters
Some interpreters encapsulate multiple internal phases to express a single, higher-level intent.
These interpreters are used when:
- A user intent can’t be represented as a single engine interaction
- Multiple steps must be coordinated safely
- Complexity should be hidden from the user
Typical characteristics:
- Internal sequencing of actions
- Unified outcome reporting
- Strong abstraction over engine behavior
- Explicit constraints to prevent misuse
Although these interpreters may perform multiple operations internally, they still represent one job step and one intent.
Advanced or escape hatch interpreters
In rare cases, interpreters may intentionally provide broader or less constrained access to an underlying system.
These interpreters exist to:
- Support legacy workflows
- Enable expert-level control
- Provide flexibility where constraint is impractical
Typical characteristics:
- Reduced guarantees compared to standard interpreters
- Explicitly documented risks and limitations
- Opt-in usage
- Clear signaling that behavior may be less stable
Escape hatch interpreters must not be treated as the default model and should coexist with simpler, intent-driven interpreters where possible.