Improve SDK first-run proof

[bmdhodl]

Summary

• make the raw quickstart generated by agentguard quickstart --framework raw --write visibly prove local budget and loop guard stops
• show guard.* events in agentguard report text and JSON output
• add focused tests and sync the README/PyPI README first-run proof path

Proof

• python -m pytest sdk/tests/test_doctor.py sdk/tests/test_demo.py sdk/tests/test_quickstart.py sdk/tests/test_cli_report.py sdk/tests/test_pypi_readme_sync.py -v -> 40 passed
• clean venv editable checkout proof confirmed agentguard.__file__=K:\agent47\sdk\agentguard\__init__.py
• clean venv ran agentguard doctor, agentguard demo, agentguard quickstart --framework raw --write, python agentguard_raw_quickstart.py, agentguard report .agentguard\traces.jsonl
• clean venv report showed Total events: 14, Estimated cost: $5.0200, guard.budget_exceeded: 1, guard.loop_detected: 1
• full SDK validation: ruff passed, full pytest passed (742 passed, coverage 92.91%), architecture tests passed (9 passed), bandit passed, release guard passed, SDK preflight passed

Scope

SDK-only. No dashboard, MCP server, registry, Glama, or marketing surfaces changed.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9db8c5b974

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Copilot]

Pull request overview

This PR strengthens AgentGuard’s “first-run” experience by making the raw quickstart visibly demonstrate local guard stops (budget + loop), extending agentguard report to surface guard event counts (text + JSON), and updating docs to match the improved proof path.

Changes:

• Update raw quickstart snippet to deliberately trigger and catch BudgetExceeded / LoopDetected, emitting guard.* events and printing a clear local-proof message.
• Extend agentguard report output (text + JSON) with a guard_events breakdown and add focused tests covering the new behavior.
• Refresh README / PyPI README and skill metadata to reflect the updated first-run proof + canonical skill location.

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
skills/agentguard/SKILL.md	Updates skill metadata version to match shipped SDK version.
sdk/agentguard/quickstart.py	Adjusts raw quickstart generator to emit/print local budget + loop guard proof and guide users to agentguard report.
sdk/agentguard/cli.py	Adds guard event aggregation to agentguard report text + JSON output.
sdk/tests/test_quickstart.py	Adds an integration-style test that runs the generated raw quickstart and validates emitted guard events + report output.
sdk/tests/test_cli_report.py	Adds coverage asserting guard events appear in report text and JSON payload.
README.md	Updates install + “Local Proof” flow to include --write, running the generated script, and reporting guard events.
sdk/PYPI_README.md	Regenerated PyPI README to stay in sync with README/CHANGELOG and new local-proof flow.
llms.txt	Updates canonical SKILL.md link to skills/agentguard/SKILL.md.
proof/skill-distribution-2026-05-08/PR_DRAFT.md	Adds PR draft proof notes for skill distribution and doc updates.
proof/skill-distribution-2026-05-08/MORNING_REPORT.md	Adds a morning report summarizing verification steps and outcomes.