Dashboard Wiki Agent Chokepoints
Agent Chokepoints
Use this before future agents change docs, runtime, generator, signing, workflow, or release behavior.
1. Technical Identity Drift
Field Content Symptom Package/feed/CLI names drift away from win11_release_guard. History / what went wrong Previous identities and prototype entrypoints can be accidentally reintroduced. Resolution / keep it this way Technical identifiers stay win11_release_guard; display name is prose-only. Do not Rename import package, feed paths, console script, JSON identity, or workflow identifiers. Verify python tools/check_project_identity.py
2. Local Display Labels Overriding Build Evidence
Field Content Symptom Local marketing labels decide installed Windows identity. History / what went wrong Some Windows 11 machines can expose stale labels while build family is current. Resolution / keep it this way Build-family and signed policy mapping drive evaluation; raw labels stay visible. Do not Let display labels override RtlGetVersion, DISM, kernel, registry, WMI/CIM build signals. Verify pytest -q tests/test_local_state.py tests/test_evaluator.py tests/test_edge_cases.py
3. WUA Treated As Verdict Authority
Field Content Symptom WUA offer/history changes the target or baseline decision. History / what went wrong WUA is localized, policy-managed, staged, and noisy. Resolution / keep it this way WUA stays read-only diagnostic context. Do not Replace signed policy target with WUA offers or history. Verify pytest -q tests/test_wua_probe.py tests/test_wua_diagnostics.py tests/test_evaluator.py
4. Special Release Becomes Existing-Device Target
Field Content Symptom 26H1 is selected for existing 24H2/25H2 devices. History / what went wrong Highest release string is not always the broad-fleet target. Resolution / keep it this way Existing-device target selection excludes special/new-devices-only releases. Do not Pick target by highest version string alone. Verify pytest -q tests/test_remote_policy.py tests/test_policy_generator.py tests/test_evaluator.py
5. Strict Production Goes Green From Fallback
Field Content Symptom Cache or bundled policy returns production-green in strict mode. History / what went wrong Fallbacks are useful but degraded. Resolution / keep it this way Strict mode needs fresh live signed remote JSON. Do not Hide fallback source status or candidate status. Verify pytest -q tests/test_runtime_policy_sources.py tests/test_cli.py
6. Public API Alias Break
Field Content Symptom /api/v1 files are missing, mismatched, or undocumented.History / what went wrong Integrations rely on stable public aliases. Resolution / keep it this way Keep v1 paths and add fields compatibly. Do not Remove v1 aliases without documented last-resort trust break. Verify python -m win11_release_guard --check-public-pages
Common Agent Mistakes Checklist
| Check | | --- | | Did not edit code when asked for docs only. | | Did not use handover files as source truth. | | Did not hide raw admin diagnostic values. | | Did not weaken tests to match a preferred narrative. | | Did not add external dashboard dependencies. | | Did not claim live checks without running them. |
Required Smoke Tests
python -m compileall -q win11_release_guard tools
pytest -q
python tools/check_project_identity.py
python tools/check_version_consistency.py
python -m win11_release_guard --self-test
Related Pages
Home | Troubleshooting | Build, Test and Release