Verification — issue-352
The testing plan's rows, executed. Commands as run; outputs summarised, with the full numbers from the final
make checkon the PR head.
Rows
| Row | Command | Result |
|---|---|---|
| T1 | pytest cli/tests/test_graph_extensions.py | green |
| T2 | pytest cli/tests/test_critics.py | green |
| T3 | pytest cli/tests/test_core_repo.py cli/tests/test_cli.py | green |
| T4 | pytest cli/tests/test_instructions.py | green |
| T5 | pytest cli/tests/test_graphlink.py | green |
| T6 | pytest cli/tests/test_graph_contribution.py cli/tests/test_graph_review.py cli/tests/test_graph_loops.py cli/tests/test_graph_refs.py | green |
| T7 | pytest cli/tests/test_migrations.py | green |
| T8 | the seven integration files named in the plan | green |
| T9 | pytest cli/tests/test_config_schema_parity.py cli/tests/test_docs_parity.py cli/tests/test_configschema.py cli/tests/test_manifest_schemas.py and uv run python scripts/validate_config.py | green; every config VALID |
| T10 | the abuse-case tests named in security-review.md | green |
| T11 | make check; grep -rn "harness_config|harness-config.default|repoInitialized|allow_repo_hooks|\.adopt(" cli/the_loop; a second grep over every shipped file for a dotted reference to a removed policy key (autonomy., security.review, tokenEconomy., userInteraction., externalTools., …) | green; the first grep is empty; the second finds only two past-tense history sentences in docs/capabilities/writing-style.md; a third grep for the third review's keys (repository., tooling., hooks., observability.) is empty |
The whole suite
Final make check on the PR head (2026-09-12, after the third review's pass):
| Check | Result |
|---|---|
ruff check cli hooks | All checks passed |
ruff format --check cli hooks | no drift |
pyright | 0 errors, 0 warnings |
markdownlint-cli2 "**/*.md" | 1086 files, 0 errors |
scripts/validate_config.py | every config validates against its schema |
pytest | 3501 passed, 1 skipped (the schema-read length-limit test in test_writing_parity is gone with the key it read; nine review-policy tests joined) |
Red first
Before the code changed, the removed reader's tests and the new assertions failed as expected: test_harness_config.py (deleted with the module), the graph-extension tests on the routing.graph.hooks shape, test_core_repo's "a committed critic is inert", and test_graphlink's "the repository's harness config is not consulted" all went red against 6bbd1e5 (the base) and green on the PR head.