Terms describing the local artifact inventory, integrity validation, and drift detection system.
Artifact
Definition
Any file produced or consumed during the mission lifecycle (e.g., spec.md, plan.md, tasks.md, prompt files). Artifacts are identified by content hash and classified by role.
Local artifact inventory and integrity validation system. It answers three questions: (1) "What artifacts exist for this mission?" — scans a feature directory, catalogs every file as an ArtifactRef with a content hash. (2) "Are all required artifacts present?" — checks against a manifest of expected artifacts per mission type/step. (3) "Have artifacts changed since I last looked?" — compares a deterministic parity hash against a cached baseline.
A checksum of the artifact's contents (using SHA256) that locks in what the artifact contains. Any changes to the artifact will cause the calculated hash to change as well, making it useful for detecting modifications.
Checksum computed from all artifact content hashes in a dossier, sorted to guarantee the same result regardless of scan order. If any artifact's content changes, the parity hash changes too. Used to answer "has anything in this dossier changed?"
A saved parity hash representing the last accepted state of a dossier. Scoped to a specific project, feature, branch, and mission so that baselines from different contexts can never be confused with each other. Drift is detected by comparing the current parity hash against this baseline.
The combination of identifiers (project, node, feature, branch, mission, manifest version) that uniquely scopes a parity baseline. Ensures that a baseline captured in one context is never accidentally compared against artifacts from a different context.
When the current parity hash no longer matches the saved baseline, something in the dossier has changed since it was last accepted. This mismatch is called parity drift. It tells the curator "artifacts have been modified — review what changed."
A frozen picture of a dossier at a specific moment: which artifacts are present, how many are required vs optional, whether all required artifacts exist, and the parity hash at that time. Scanning the same unchanged content always produces the same snapshot.
The process of walking through a feature directory, identifying each artifact file, computing its content hash, and recording it in the dossier. Also checks for any required artifacts that are missing.
When artifact indexing finds that a required artifact is absent or unreadable, it records the gap with a reason code explaining why: not_found (file doesn't exist), unreadable (file exists but can't be read), invalid_format (file exists but isn't valid), or deleted_after_scan (file disappeared during indexing).