Changelog

0.2.0 — 2026-09-28

Changes from the first end-to-end run on a real repository:

• Playwright is found in the global npm root when the project doesn't depend on it, and doctor reports Playwright, its Chromium, and gh --attach support.
• New rehearse command: dry-run an unsealed scenario against the unfixed app. Rehearsals are stored apart from captures and can never be registered as evidence.
• New media --format still-pair for labeled before/after screenshot pairs.
• Fixed: side-by-side videos never actually padded the shorter clip (filter order), and the "final frame held" label covered the whole clip. The note now shows only while held.
• Uploaded videos are embedded as a bare URL line so GitHub renders a player.
• status and report print a compact summary; status --full prints everything.
• init writes every summary.json field, now documented in the configuration reference.
• SKILL.md states the ordering rules the helpers enforce (branch before init, baseline before regression tests, fresh dev server per capture) and adds pacing guidance.
• Packaged as the implement-and-prove plugin (previously proof) in the multi-skill NotHassan/skills repository. The plugin command is /implement-and-prove:implement-and-prove.

0.1.0 — 2026-09-26

• Portable Agent Skill with Claude marketplace/plugin metadata and Codex UI policy.
• Safe user/project installer, explicit invocation, progressive reference loading.
• Sealed acceptance/scenario contracts; repeatable Playwright before/after capture.
• Static-command verification, runtime events, and stale-evidence checks.
• FFmpeg H.264 walkthroughs and labeled side-by-side/sequential comparisons.
• Reviewed evidence registry, managed PR body, draft-only GitHub publication.
• Runtime --attach detection and partial-upload reconciliation.
• Unit/media tests and an optional real-browser smoke harness.

Initial implementation. Live authenticated publication and native host installation must be validated separately; the repository does not claim those tests ran here.