Deferred entirely, not attempted this pass

Index: Known gaps

Status (re-verified 2026-10-09 at 456989f): POLICY_NOTE. Deferred-scope note. .devcontainer/ still does not exist (ls fails).

(heavier structural changes, each needing its own scoped design rather than a drive-by addition): - Devcontainer image — a maintained .devcontainer/ needs a decision on which system tools (castxml, libabigail, abi-compliance-checker, compilers) ship baked-in vs. installed on first use, and upkeep as those pins drift; pixi (see CONTRIBUTING.md) already solves the "one command gets you a working dev environment" problem this would target, without the image-maintenance burden. - Trend-reporting database — persisting scripts/check_tier_accuracy.py /check_fp_rate.py/mutation-score history across runs (rather than each CI run only gating against a static baseline) needs a storage decision (artifact-based vs. external DB) and a retention/access policy before it's worth building. - Full behavioral baseline — skills-src/evaluation/agents/ (this pass, M1-5) is a real but minimal harness with one task; a "full behavioral baseline" implies a broad task suite plus a scoring/leaderboard story, which should grow from real usage of the one-task harness rather than being speculatively built out now.