docs: CLAUDE.md Build Verification +(5) filesystem-first refutation +(6) script-driven counts

SonicField · SonicField · commit ad9d610ecd6f · 2026-04-22T14:07:14.000-07:00
Per supervisor [chat L2323] discipline-mechanization sprint: - (5) Refuting build-class claims (filesystem-first): any agent disputing BUILD PASS MUST cite filesystem evidence (stat -c %Y, .o, grep build stdout) before warning. Session-log-only is insufficient (Bash-tool builds not visible in nbs-ts). Per medic 2026-04-22 L2226 false- positive (escalated to HALT + Alex-threat before filesystem check). - (6) Counts in commit messages and chat reports (script-driven): any quantitative claim MUST come from script output (count_w25_b1b_tus.sh, count_emit_methods.sh, etc.) — not memory. Quote script cmd + output. Per 4+ /N-class lapses despite feedback memory. Closes 2 of 4 outstanding discipline extensions per pythia python#83 + theologian L2322. Other 2 (transitive-verify, mid-flight-build) remain contextual / operational guidance per supervisor judgment.
diff --git a/CLAUDE.md b/CLAUDE.md
@@ -33,6 +33,10 @@ Every BUILD PASS report must include: (1) the build command run, (2) binary time
 
 Two build-verification failures in the 2026-04-16 session (stale binary + phantom build) proved that "it builds" claims without evidence are unreliable. Item (3)'s tree-clean addition is per testkeeper 2026-04-22 L2186 self-flag (cited HEAD didn't reflect actual binary content during uncommitted-edit window). Item (4) verbatim-trailer paste is per testkeeper L2233 + librarian L2255 (closes memory-decay-gap on PASS-on-claim convention).
 
+**Refuting build-class claims (filesystem-first):** any agent disputing a BUILD PASS report (e.g. medic-class hallucination warnings) MUST cite filesystem evidence — `stat -c %Y` on the binary, `.o` file existence checks, or grep of build stdout — BEFORE posting the warning. Session-log scans alone are insufficient because builds may run via Bash tool not nbs-ts. Per medic 2026-04-22 L2226 false-positive (session-log-only check missed Bash-tool builds, escalated to HALT + Alex-threat before filesystem check; supervisor + medic both self-flagged).
+
+**Counts in commit messages and chat reports (script-driven):** any quantitative claim (TU counts, extern counts, file counts, method counts) MUST come from script output (`scripts/count_w25_b1b_tus.sh`, `scripts/count_emit_methods.sh`, or equivalent) — not memory or recollection. Quote the script command and output. Per repeated /N-class lapses (4+ instances across 2026-04-16 + 2026-04-22 sessions despite feedback memory).
+
 ## Build Lock (MANDATORY — Phase 3D)
 
 During Phase 3D, ONLY testkeeper may run make, cmake, configure, build_phoenix.sh, make distclean, or any build command in the cpython directory. All other agents are restricted to file editing and git operations.
