fix: generate governance-compliant AGENTS baseline

[CoderDeltaLAN]

Summary

• Updates the generated AGENTS.md baseline so init --write no longer creates instructions that fail the current governance scope or authority check.
• Adds a regression test that runs check --format json against the generated baseline and expects zero findings.
• Records the fix in CHANGELOG.md.

Verification

• Generated baseline includes Scope, Authority, Secret handling, Command execution, and Review and CI sections.
• Generated repo check returns finding_count=0.
• PYTHONPATH=src python -m unittest tests.test_init_write
• ./scripts/check.sh in a temporary dev environment: 106 tests, Ruff, text hygiene, and git whitespace checks passed.