DoneAudit v0.1.1 — Marketplace / distribution release

Marketplace / distribution release

• Rename the public GitHub Action to DoneAudit - Verify AI Completion and clarify its description.
• Add a copyable kafidog/DoneAudit@v0.1.1 workflow example.
• Retain the upstream receipt gate, internal formats, and MIT attribution unchanged.

The npm/Codex CLI remains doneaudit@0.1.0; this release changes Action distribution and documentation, with no new verification functionality. GitHub Marketplace listing is now live: https://github.com/marketplace/actions/doneaudit-verify-ai-completion

Validation: all six focused Action self-tests passed, including fresh success, deliberately failing checks, and missing-evidence rejection.
https://github.com/kafidog/DoneAudit/actions/runs/34171473872

DoneAudit v0.1.0 — Saying done does not make it done

DoneAudit v0.1.0 turns an AI completion claim into an evidence check: real command receipts, a deterministic 0–100 score, passed/failed/insufficient items, and the same result in GitHub Actions.

Install from your Git project:

npx --yes --package=https://github.com/kafidog/DoneAudit/archive/refs/tags/v0.1.0.tar.gz doneaudit init

Then ask Codex to do the work and follow the installed completion rule. Review the detected check commands once; missing commands never count as passing.

Validation: real Codex CLI workflow on Windows; fresh packed and public-source installation; intentional test failures blocked; success=100/VERIFIED and failure=70/FAILED; GitHub Windows/Linux release gate; preserved upstream compatibility tests.

The score measures evidence coverage, not the probability that a program is correct. The Codex integration uses AGENTS.md instructions; it is not a hard stop hook. GitHub reruns the checks independently.

Built on mohamedzhioua/agent-done-or-not (MIT, Copyright (c) 2026 Zhioua Mohamed). Original license, authorship, gate engines and Git history are retained. See THIRD_PARTY_NOTICES.md.

• Real terminal demo
• Windows/Linux release check
• Latest handoff