Changes

1.0.0 — 2026-10-22

• Completed the owner-delegated AI design review and added a coherent final Northstar example, with a narrowed evidence-interpretation claim, explicit practical scoring, twelve item-review records and nine worked scoring responses.
• Resolved weighting, timing, capacity and study-sequence choices. The six-person feasibility design has no passing score or credential effect; real organizational authority and candidate evidence are not invented.
• Retained original eval artifacts and the earlier candidate example without transferring their approvals or scores. Runtime skills/helper are unchanged from the observed native create/resume candidate.
• Finalized the v1 distribution and promotion copy with an explicit Windows/Python 3.11 verification claim. Cross-platform CI is configured but unobserved; Claude activation remains unverified.

1.0.0-rc.1 — 2026-10-22

• Prepared a public release candidate with an allowlisted sanitized export, original/public file hashes, and tests for disclosure and overwrite boundaries.
• Added offline CI for Windows, macOS and Linux and a v1 command/record compatibility policy. Hosted CI execution remains unobserved until repository publication.
• Added an owner-review packet with all twelve Northstar items, actual program artifacts and five domain decisions. Revised nine selected-response items to reduce wording cues and split the access practical's overloaded scope criterion. All remain drafts; historical eval outputs and pilot scores were not changed or transferred.
• Clarified the item skill's authoring review and actual helper commands. The ten-skill scope and record schema remain unchanged.
• Final 1.0.0 is pending the gates recorded in docs/release-candidate.md.

0.1.2 — 2026-10-22

• Added the two-session Northstar program-owner evaluation: creation from raw sources, pilot interpretation, product change and release recommendation.
• Added preserved receipts, output snapshots, separate artifact reviews and grader negative controls.
• Made source preservation, historical analysis and verified handoff/restore explicit in the skills.
• Reworked onboarding around the owner journey and added security/contributor guidance.
• See the verification report for executed results and remaining limits.

0.1.1

Bound approvals and scored forms to exact content, policies, evidence and award intent. Added CLI scenarios and boundary regressions. Historical findings remain in docs/evaluation-findings.md.

0.1.0

Initial portable ten-skill lifecycle package, local helpers, synthetic Harbor example and isolated installer.