Pull request overview

This PR adds a new Python agent skill that documents the end-to-end workflow for preparing Python package release PRs in the microsoft/agent-framework monorepo, and wires it into the Python agent documentation index.

Changes:

• Add agent-framework-py-release skill describing CHANGELOG-driven version bumping, lifecycle tier rules, and dependency-floor validation.
• Update python/AGENTS.md to list the new release skill alongside existing Python skills.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 10 comments.

Automated Code Review

Reviewers: 4 | Confidence: 95%

✓ Correctness

This PR adds a new Python release skill document and a corresponding reference in AGENTS.md. Both changes are documentation-only. The skill file references real, verified repo infrastructure (PACKAGE_STATUS.md, validate-dependency-bounds-test poe task, the root pyproject.toml core pin, and the validator script). The AGENTS.md entry follows the existing format. No correctness issues found.

✓ Security Reliability

This PR adds a new AI agent skill document (SKILL.md) with release workflow instructions and updates AGENTS.md to reference it. The content is purely instructional documentation for AI coding agents — no executable code, no CI pipeline changes, no secrets, and no trust-boundary interactions. The shell commands shown are illustrative templates, not auto-executed scripts. No security or reliability issues identified.

✓ Test Coverage

This PR adds a new documentation-only skill file (SKILL.md) for Python release workflows and a one-line reference in AGENTS.md. No production code or behavioral logic is changed. The existing .github/skills/ directory already contains six other skill files (python-development, python-testing, python-code-quality, python-feature-lifecycle, python-package-management, python-samples), none of which have dedicated tests for their content or consistency with AGENTS.md. The core skills framework (FileSkill parsing, discovery) is well-tested in python/packages/core/tests/core/test_skills.py, but those tests cover the skill-loading machinery, not individual skill document content. Since this PR follows the established pattern for adding skills and introduces no executable code, there is no test coverage gap specific to this change.

✓ Design Approach

The new release skill is close, but its step 4a reconciliation workflow has two design flaws that make the changelog-driven bump check unreliable: the sample awk extraction does not actually capture the drafted section body, and the reconciliation logic still excludes root agent-framework changes even though the skill says root changes can drive an independent bump.

Automated review by moonbox3's agents