Isolate skill evaluations and prepare v1 release

[dd3ok]

Summary

• execute evaluation workspaces outside the repository and reject known same-name skill discovery ancestors
• require protocol v3 adapter isolation attestations and archive only bounded report artifacts
• document pinned v1.0.0 installation and the isolated evaluation contract in English and Korean

Why

A report path under the repository could make no-skill baselines discover the repository's .agents or .claude mirrors. Separating execution from reporting and requiring adapter attestation prevents those contaminated runs from being accepted.

Verification

• python scripts/verify_all.py on Windows: 81 tests, 1 POSIX-only skip
• python3 scripts/verify_all.py in WSL with Python 3.14.4: 81 tests
• python -m ruff check .
• 36-run fixture smoke with report output under reports/ and auto-cleaned work root outside the repository
• gh skill publish --dry-run .

[gemini-code-assist]

Code Review

This pull request updates the evaluation harness to protocol version 3, introducing isolated execution workspaces to prevent skill contamination and requiring adapters to return an isolation attestation. It adds a --work-root option, resolves runner command file arguments to absolute paths, and validates that no same-name skills exist in ancestor paths. The feedback highlights an issue in prepare_work_root where a newly created, user-specified work root is not cleaned up if subsequent validation checks fail, and suggests tracking directory creation to ensure proper cleanup.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.