Align ADR authoring and decision-impact guidance with the ratified policy #91

Outcome

Bring Aether's existing ADR authoring skill and managed decision-impact guidance into alignment with the ratified Hygiene ADR policy so repository agents can reconstruct historical decisions and keep them current during normal issue/PR work.

This is a focused follow-up to completed #49. Reuse and upgrade the existing skill and managed module; keep the broader repository-default bundle in #67.

Verified gap — 2026-09-26

• library/organization/projections/templates/decision-impact.AGENTS.md still inherits Hygiene ADR policy v1.0.0 at 5e0602265b6ac5e5165b89f418e55a3fd12f8a64 with proposed authority.
• Hygiene feat(hooks): harden opt-in agent hooks and disposition staged prototypes #15 is already closed and ratified. Its v1.1.0 policy and ratification evidence are present at c589587395750cd1c79c6fa0bef010189c547249. Verify the supported immutable source matrix during implementation; do not request the same ratification again.
• library/organization/skills/architecture/create-decisions-document/SKILL.md exists, but centers DECISIONS.md and excludes unaccepted proposals, while the managed hook explicitly routes consequential new choices to proposed ADRs.
• Existing repository backfill issues already require evidence-backed historical reconstruction, canonical docs/decisions/ adoption, and continuous capture. They need one consistent reusable authoring path.

Required work

1. Reconcile the canonical decisions specification, skill, templates, references, evaluations, managed hook, and generated projections against the accepted Hygiene contract. Preserve Aether lifecycle/version semantics and record exact upstream policy revisions.
2. Support the distinct operations: historical reconstruction, new proposed ADR, evidence/outcome correction, reference to an existing decision, and proposed supersession. Make the canonical index and legacy DECISIONS.md navigation behavior agree with the supported Hygiene/Holon migration contract.
3. Define historical reconstruction from reachable Git history, merged PRs, issues, releases, architecture documents, and existing records. Group evidence by consequential decision; preserve IDs, source links, contemporary rationale, recorded dates, and gaps. Record the reconstruction date separately from any proven historical decision date.
4. Keep decision disposition, implementation, and verification independent. A merge or implementation proves neither rationale nor human acceptance. Unsupported historical status remains proposed or explicitly unresolved under the policy.
5. Generate concise managed AGENTS.md guidance that routes to the shared skill and performs a decision-impact check before implementation and again before issue completion/PR handoff. Preserve consumer-authored instructions.
6. Require PR handoff to reference the governing/new/updated ADR or state ADR not required with a concise reason. Consequential changes need an appropriate record; routine implementation under an existing design needs no duplicate ADR.
7. Document preview, pinned adoption, upgrade, missing-skill behavior, validation, and rollback through existing Aether/Holon mechanisms. Instructions route agent behavior; do not claim that adding an AGENTS block alone enforces every PR.

Acceptance criteria

• The skill, governing spec, templates, references, and hook consistently support the accepted policy and proposed-record authoring.
• Exact policy references and authority evidence are recorded without silently promoting unrelated draft artifacts or inherited contracts.
• Historical reconstruction preserves source evidence, existing identity/lineage, uncertainty, and historical versus reconstruction dates.
• Evaluations cover a well-supported historical record, missing rationale/authority, a new proposal, supersession, routine implementation, and legacy index migration.
• Generated guidance routes create/update/supersede/reference/no-ADR cases and preserves repository-local content.
• PR handoff examples contain an ADR reference or justified no-ADR disposition.
• Repository-native validation and skill evaluations run locally; hosted/host-specific evidence remains explicit if unavailable.
• Documentation identifies the pinned adoption path for the first repository canary.

Dependencies and coordination

• Policy authority: Define the global Architecture Decision Record standard and inheritance contract hygiene#15
• Scaffolding/materialization compatibility: Add ADR scaffolding and inherited repository decision policy holon#6
• Delivered hook: Add a decision-impact and ADR hook to canonical AGENTS guidance #49; broader default-bundle integration: Publish a repository-default skill bundle and generated AGENTS guidance #67
• Validation interface: Deliver #5 through an ordered architecture-validation execution plan relay#99 and Validate ADR, roadmap, and Repository Intelligence linkage egolint#24
• ADR/Decisions campaign: [Capability 1] Roll out and reconcile the ADR baseline and Decisions views across the fleet pace#5
• First planned consumer: Backfill repository ADR history and enable continuous decision capture identity#69

This authoring update can be developed against the documented validation boundary; it does not require the complete fleet rollout or closure of #67.

Non-goals

Backfilling Aether's own history (#85), bulk editing consumer repositories, inventing historical reasoning or authority, implementing an ADR validator in Aether, mandatory ADRs for every edit, autonomous acceptance, or deployment.