|
| 1 | +# Drift Monitoring Plan |
| 2 | + |
| 3 | +**Status**: DRAFT / PLAN ONLY |
| 4 | +**Scope**: how Riverbraid may detect and report drift |
| 5 | +**Implementation status**: NOT_IMPLEMENTED |
| 6 | + |
| 7 | +## Purpose |
| 8 | + |
| 9 | +Drift monitoring is the process of detecting when a repository, verifier, registry entry, claim, or evidence surface moves away from its declared state. |
| 10 | + |
| 11 | +The goal is not to prevent all change. The goal is to make unaccounted-for change visible. |
| 12 | + |
| 13 | +## Drift categories |
| 14 | + |
| 15 | +| Drift type | Description | Example | |
| 16 | +|---|---|---| |
| 17 | +| Claim drift | Documentation claims more than evidence supports. | A README implies production readiness. | |
| 18 | +| Registry drift | Registry pin no longer reflects intended proof surface. | Refusal-Gold has a later fail-closed patch not reflected in the pinned registry. | |
| 19 | +| Verifier-depth drift | A verifier's actual depth differs from declared depth. | Petal says deterministic-vector check but only checks file presence. | |
| 20 | +| Workflow drift | CI path changes without evidence update. | Verification command changed but evidence ledger not updated. | |
| 21 | +| Dependency drift | Dependencies change without boundary update. | Lockfile regenerated without recorded evidence. | |
| 22 | +| Artifact drift | Generated files, logs, or evidence packets diverge from expected state. | Evidence packet schema changes without version bump. | |
| 23 | +| Governance drift | Authority or decision process changes without recording. | Registry pin updated outside succession rules. | |
| 24 | + |
| 25 | +## Detection sources |
| 26 | + |
| 27 | +Potential sources: |
| 28 | + |
| 29 | +- GitHub compare results |
| 30 | +- GitHub Actions metadata |
| 31 | +- Evidence packets |
| 32 | +- claim ledger |
| 33 | +- evidence ledger |
| 34 | +- readiness matrix |
| 35 | +- registry scope gap |
| 36 | +- package lockfiles |
| 37 | +- workflow files |
| 38 | +- petal manifests |
| 39 | + |
| 40 | +## Drift record format |
| 41 | + |
| 42 | +```yaml |
| 43 | +drift_id: |
| 44 | +detected_at: |
| 45 | +repository: |
| 46 | +drift_type: |
| 47 | +observed_change: |
| 48 | +expected_state: |
| 49 | +evidence_source: |
| 50 | +severity: |
| 51 | +status: |
| 52 | +assigned_issue: |
| 53 | +resolution: |
| 54 | +non_claims: |
| 55 | +``` |
| 56 | +
|
| 57 | +Allowed statuses: |
| 58 | +
|
| 59 | +```text |
| 60 | +OPEN |
| 61 | +UNDER_REVIEW |
| 62 | +ACCEPTED |
| 63 | +FALSE_POSITIVE |
| 64 | +RESOLVED |
| 65 | +DEFERRED |
| 66 | +``` |
| 67 | + |
| 68 | +## Severity guide |
| 69 | + |
| 70 | +| Severity | Meaning | |
| 71 | +|---|---| |
| 72 | +| Low | Documentation or metadata drift that does not affect proof claims. | |
| 73 | +| Medium | Drift that may affect claim clarity or reviewability. | |
| 74 | +| High | Drift that affects verification behavior or registry truth. | |
| 75 | +| Critical | Drift that may create a false public trust signal. | |
| 76 | + |
| 77 | +## First implementation target |
| 78 | + |
| 79 | +The first drift monitor should be static and report-only: |
| 80 | + |
| 81 | +1. Compare tracked documentation claims against known non-claims. |
| 82 | +2. Compare active repo list against registry list. |
| 83 | +3. Compare declared verifier depth against known classification. |
| 84 | +4. Report differences. |
| 85 | +5. Open no automatic PRs. |
| 86 | + |
| 87 | +## Fail-closed rule |
| 88 | + |
| 89 | +If drift affects a proof claim, the affected claim should be downgraded to `UNKNOWN_PENDING_EVIDENCE` until reviewed. |
| 90 | + |
| 91 | +## Non-claim |
| 92 | + |
| 93 | +This plan does not implement drift monitoring and does not claim real-time monitoring exists. |
0 commit comments