100%

Section 06 — verify

Goal

Prove the functional mission: a production-path run finds substantive findings and every rule is provably checked.

Implementation Sketch

Author the full test suite (parser, schema-invariant, state machine, batch dispatch/router, coverage assertion). Add a coverage assertion that every PREFIX-NN rule is routed to exactly one batch and reached. Run the rewritten skill end-to-end on a real code change and confirm it surfaces non-trivial, multi-category findings the current engine misses. --self-test green end-to-end.

Spec References

CLAUDE.md §Fix Completeness (production-path end-to-end); script-first.md §4 (tests mandatory); layer-coverage N/A (non-compiler).

Work Items

  • Author the full pytest suite + the every-rule-routed-and-reached coverage assertion
  • Run the rewritten skill end-to-end on a real diff; confirm substantive multi-category findings vs the old engine
  • --self-test green end-to-end; record the production-path evidence

Fresh intel (regenerated)

{ “schema_version”: 1, “target”: { “kind”: “plan-section”, “ref”: “impl-hygiene-rewrite-v3/s-34875c2a” }, “generated_at”: “2026-06-23T06:00:33.967526+00:00”, “graph_state”: { “head_sha”: “ddf60faa”, “last_code_import_at”: “2026-06-23T05:57:12.447Z”, “embedding_stale”: false, “insights_stale”: false, “cpg_stale”: false }, “surfaces”: {}, “summary”: “intel-package[plan-section:impl-hygiene-rewrite-v3/s-34875c2a] agent-authored dossier”, “degraded”: false, “dossier”: { “objective_symbols”: [], “tiers”: {}, “agent_authored”: true, “difficulty”: { “class”: “routine”, “research_online”: false, “research_mode”: “auto”, “signals”: [], “plan_dir”: “/home/eric/projects/ori_lang/plans/impl-hygiene-rewrite-v3” } } }

(full dossier: verify—s-34875c2a.intel.json)

HISTORY

  • 2026-06-23 — w-283e7945 coverage assertion landed; verify-finish scope mapped: added test_coverage_assertion.py (4 tests) — the every-PREFIX-NN-rule-routed-to-exactly-one-batch-AND-reached regression pin (sc-0015 coverage half; the BUG-07-305 functional-negative pin): 460 rules partitioned over 5 batches, no drop/dup, all 13 categories covered, the lint->B1..B5->done walk reaches every batch, every reached rule tier-classified. REMAINING (deliberate careful finish): (1) w-30d0d504 —self-test green — BLOCKED by hanging seam_scan + detector tests (test_seam_scan + test__detectors exceed 2m for 2 files — graph-I/O hang, pre-existing per the recon dossier). Those test the DEAD seam_scan/detector pipeline (unused by the live rails path — only orphaned legacy phase-2-landscape.md / phase-3-analysis.md docs reference seam_scan/batch_routing). Clean fix: remove the dead seam_scan + detector modules + their tests + the orphaned phase-.md docs + the legacy cmds (seam-scan/set-scope/record-synthesis/start/set-target) — PRESERVING the LIVE finding_emit + verification_gate (used by cmd_emit_findings + scripts/plan_orchestrator/tests/test_findings_loop_parses_all_runtime_emissions.py) + the rails core. (2) w-e5a61cbb — run the rewritten /impl-hygiene-review end-to-end (rails-start -> per-batch subagents over a real diff -> inline-fix) confirming substantive multi-category findings vs the old engine (the sc-0014 / plan-wide sc-0001 functional proof). The finish is a large careful pass (dead-pipeline gut distinguishing live finding_emit/verification_gate + an integrated multi-subagent review run), executed deliberately per CLAUDE.md NO-SHORTCUTS, not rushed.