Pilot optional diagnosis records across the repository fleet

[szmyty]

Outcome

Pilot the diagnosis-record capability in a small, representative set of Ego Hygiene repositories, measure authoring and maintenance friction, and prepare a review-gated fleet rollout only after the contract, policy, validation, and projection path are ready.

Dependency chain

• Aether artifact and skill: Define evidence-labelled diagnosis records and authoring skill aether#74
• Hygiene applicability: Define repository applicability for diagnosis records hygiene#41
• Egolint validation: Validate repository diagnosis records and derived indexes egolint#50
• Observatory read model: Model diagnosis records in Repository Intelligence observatory#16
• Relay diagnostics view: Add a diagnostics ledger to Repository Intelligence relay#57

Pilot design

Select at least:

• one small public repository;
• one repository with a meaningful historical debugging record;
• one private or restricted-data scenario exercised only with synthetic evidence.

For each pilot:

• declare capability state and pinned contract;
• add the bounded record path without forced historical backfill;
• create or migrate one real, reviewable record where publication is safe;
• validate with Egolint;
• verify Observatory projection and Relay rendering;
• record authoring time, validation friction, redaction burden, false positives, unknowns, and maintainer feedback;
• test opt-out, no-record, stale-record, restricted, and superseded behavior;
• preserve repository-specific governance and never copy private evidence into a public canary.

Rollout gates

Fleet proposals may begin only when:

• the Aether contract is accepted and version-pinned;
• Hygiene applicability is explicit;
• Egolint emits stable machine-readable results;
• Observatory and Relay preserve sensitivity and uncertainty;
• the pilot identifies rollback and removal behavior;
• maintainers review the proposed repository classes and default state.

Acceptance criteria

• Pilot repositories and selection rationale are documented.
• Adoption is opt-in and review-gated.
• No historical backfill is implied or fabricated.
• Each pilot has before/after conformance evidence and represented commits.
• Authoring, validation, redaction, maintenance, and dashboard usability findings are recorded.
• Private/restricted behavior is proven with synthetic data only.
• Removal and rollback affect only owned generated/adopted files.
• A fleet rollout proposal preserves optional, required, not-applicable, unknown, blocked, and opted-out states.
• Pace proposes changes; it does not silently remediate repositories.

Related

• Roll out the Repository Intelligence profile across eligible repositories #13
• Build the Repository Intelligence experience relay#27
• Build the normalized Repository Intelligence graph and query snapshots observatory#7

Non-goals

• Editing every repository immediately.
• Requiring a record for every defect.
• Publishing private debugging history.
• Treating record counts as repository quality scores.