-
Notifications
You must be signed in to change notification settings - Fork 0
Glossary
Keshav edited this page Jul 29, 2026
·
1 revision
This vocabulary is normative for product documentation and contracts. New terms should not duplicate an existing concept with different wording.
- Organization: The top-level tenant and policy boundary.
- Project: A user-defined engineering boundary containing repositories, services, environments, resources, and incidents.
- Repository: A Git or filesystem source root inspected by StackCendra.
- Service: A runnable or deployable unit, such as an API, worker, frontend, or database dependency.
- Environment: A named runtime context such as local, testing, staging, production, or disaster recovery.
- Scan: A bounded, read-only inspection of a trusted directory at a particular point in time.
- Detector: A deterministic, versioned component that emits facts from supported evidence.
- Fact: A normalized statement about a project, service, requirement, resource, or runtime.
- Evidence: A minimal reference supporting a fact or hypothesis, including provenance and observation time.
- Observed fact: A fact parsed directly from an authoritative source.
- Inferred fact: A fact derived deterministically from one or more observations.
- Hypothesis: A provisional explanation that may be generated heuristically or by AI and requires validation.
- Confidence: A calibrated measure of support for an inference or hypothesis. Confidence is not authorization.
- User correction: An attributable decision that replaces or qualifies a detected result.
- Configuration contract: A versioned schema describing the names, types, constraints, secrecy, defaults, and validations required by a service.
- Configuration version: An immutable instance of a configuration contract and its environment-specific references.
- Secret reference: An opaque pointer to protected credential material. It is not the secret value.
- Drift: An unintended or unvalidated difference between expected and observed environment state.
- Readiness score: A summarized deployment assessment derived from explicit blocking findings and warnings. It never overrides a blocker.
- Action: A versioned, parameterized operation with defined permissions, targets, validation, and audit behavior.
- Flow: A durable graph of actions, conditions, approvals, health gates, retries, and rollback behavior.
- Runner: A constrained execution component authorized to perform a specific action against a specific target.
- Capability: A short-lived, signed authorization granting a narrowly defined operation.
- Deployment: A traceable promotion of versioned code, artifact, and configuration state to an environment.
- Approval: An attributable decision satisfying a named policy requirement for a specific immutable operation.
- Resource: A host, container, cluster, cloud object, database, or other operational dependency.
- Telemetry: Structured logs, metrics, and traces describing runtime behavior.
- Incident: A time-bounded investigation connecting symptoms, affected resources, evidence, hypotheses, actions, and recovery.
- Remediation: A proposed or executed change intended to resolve an incident.
- Recovery verification: Deterministic checks demonstrating that health has returned to an accepted state.
- Reproduction bundle: A sanitized, isolated set of code, configuration structure, synthetic data, dependencies, and representative inputs used to recreate a failure locally.
- Local-only vault: A credential mode in which decryption occurs only on the user's device and cannot support unattended cloud-initiated sessions.
- Automation vault: A credential mode allowing narrowly authorized runner decryption through envelope encryption and complete auditing.
- Policy: A deterministic rule governing whether an operation is permitted and which approvals or safeguards it requires.
- Audit event: An immutable security-relevant record of an actor, target, operation, decision, result, and correlation context.
- Trusted root: A user-approved filesystem directory that bounds a discovery operation.
- Customer-hosted runner: A runner deployed within a customer-controlled network and connected outward to the control plane.
StackCendra is currently in Phase 0. The complete Phase 0–13 plan is versioned in the main repository under docs/wiki; later-phase pages describe intended behavior, not current implementation.
- Phase 0 foundation
- Phase 0 backlog
- Release 0.1 discovery
- Phase delivery framework
- Roadmap
- Wiki review guide
- Risks and decisions