fix(adr-authoring): align decisions with ratified Hygiene policy

[szmyty]

The ADR skill previously excluded proposals while the managed decision-impact hook required them and still selected pre-ratification policy. This aligns the existing specification, skill, templates and generated guidance with ratified Hygiene ADR policy v1.1.0 at c589587395750cd1c79c6fa0bef010189c547249.

Authors can now reconstruct historical decisions, draft proposals, append sourced corrections/outcomes, reference governing records and propose replacements. New records remain proposed; human disposition, implementation and verification stay separate. The guidance checks decision impact before implementation and again at completion, and requires an ADR reference or a justified ADR not required handoff.

• Preserves IDs, evidence gaps, historical/reconstruction dates, legacy content and consumer instructions. New canonical records use docs/decisions/; migration is reviewed and reversible.
• Records exact policy authority and source digests. Aether skill/spec/module stay draft; the independent Intelligence contract stays proposed. The projection builder rejects skill/hook policy drift.
• Documents pinned preview/adoption/upgrade/rollback for the first canary, Identity Define scoped agent-session and garden-access capabilities #69 under Pace feat(skills): complete the five foundation-document skill packages #5. Regenerates the catalog, portable skill/spec and all provider projections; refreshes root continuity.

Validation: 199 tests passed with zero skips, including all 11 new ADR tests using the pinned Hygiene source and real EgoLint/Relay runtime, plus the existing native title tests. 205 deterministic evaluation cases passed, including nine ADR examples. Repository, shell, strict staging, catalog/schema/provenance and generated-output checks passed. Two clean builds produced 334 byte-identical files; gh skill publish dist --dry-run passed without publishing. The existing preserved staging-hash provenance warning remains.

Compatibility limits are explicit in docs/adr-authoring-adoption.md: selected Holon blueprint 1.0.0 still emits an older policy pin that native validation rejects, so a reviewed owner upgrade precedes materialization. The pinned Hygiene decision-set helper disagrees with EgoLint about pending reciprocal lineage; proposals must preserve the accepted predecessor. Deterministic examples are not model/host execution evidence. Hosted architecture acceptance remains deferred under Relay #99; this does not complete fleet adoption, releases or Aether's later backfill (#85).

ADR not required: applies ratified Hygiene ADR-002 within Aether ADR-001/ADR-002/ADR-004 source, ownership and versioning boundaries; introduces no new policy authority, artifact kind or distribution channel.

Closes #91.

[devactivity-app]

Pull Request Summary by devActivity

Metrics

Achievements

• @szmyty
  Earned XP: 10⭐
  Check Your Progress