webby: row 217 follow-up on b6c3129d2; and session line

agents/webby/work/queue-217/EARLY-LOOK-c25f5b8da.md

@@ -86,3 +86,38 @@ asked. It isn't a verdict, and I ran nothing.
 
 Per the brief: the library, bus, business, cli and harness suites green, every scripts/test-*.sh
 on CLEAR, M1-M4 killed, and the obsolete-code grep. Nothing above has been run.
+
+## Follow-up, b6c3129d2 (beakley-0010, sage-3668/3670)
+
+- The tip fast-forwards from c25f5b8da with one commit and no trailer. 11 files, +212/-6: tests,
+  BUILD.md, BUILD-LOG, the manifest and mutants.py. No product line changes. Unrun.
+- E1 closed statically. reply-limits.test adds actor `mallory`, who sets `human` to `mallory` and
+  changes replyLimits. mallory isn't the tester or a role id, so only the `old.human` comparison
+  refuses it, and M2's swap lets it through. M2 isn't run, so it's not a kill yet.
+- E2 closed statically. The new citation-guard case commits the file with
+  `Mosaic-Agent: acme/scribe`, adds an unrelated commit, serves and cites that tip, and pins
+  `{scope: 'canon', agent: 'acme/scribe', file}` in the stored `message.checked` event and
+  `last: file` in thread.messages. So the file commit, not the cited tip, is what gets stored.
+- E3. I was wrong about the premise. `organization.set` exists (broker.mjs:4701/4709), but
+  ORGANIZATION_KEYS leaves out replyLimits. sage-3668/3670 accept direct writer.write under
+  the chief and tester actors, and BUILD.md now gives the right reason.
+- E4. The tests come first, and the product fix waits for the RED on CLEAR. citation-availability
+  proves a valid cite through the real mirror and socket. Then, while withLibraryLock holds, it
+  expects `citation-unavailable` with the exact reason and messages unchanged. The fixed fallback
+  still goes out, and the cite passes again after release. Malformed sync state also refuses
+  `citation-unavailable`. No mirror and a fresh empty mirror still refuse `citation-invalid`. The
+  runner case checks that `citation-unavailable: <reason>` reaches the actual retry request.
+  requests[1] is a string there, since line :55 does `assert.match` on it, so `.includes` works.
+  M5's anchor is the planned refuse line, which isn't in guard.mjs yet (:177 is still
+  `catch { served = null; }`). mutants.py refuses a missing anchor (:59), and its control would be
+  RED until the fix anyway, so M5 can't report a false kill.
+- E5 is recorded as an inherited limitation, not a rule. That's enough for me.
+- E6 closed. The guard table runs as both coder and research.
+- E7 closed statically. citation-runtime.test starts the real runtime with the real organization
+  file. It changes the file, corrupts it and removes it, and reads library.status each time,
+  without replacing the library adapter.
+- F1 (new, minor). citation-helpers now builds the guard's reader from `served` on each call, where
+  it used to be fixed at r.commit. That changes every row 217 test that calls setServed, which is
+  the intent. The adapter's `status` still answers `served: r.commit`. No test reads it after
+  setServed today, but it would read wrong if one did.
+- Still open: the E4 product fix and the RED-then-green on CLEAR. The supporting scope waits on 283.

docs/SESSIONS.md

@@ -1535,3 +1535,4 @@ are never rewritten or removed; corrections are new entries.
 2026-10-23T19:35Z | Webby (Opus 5.5, reviewer seat) | row 288 follow-up on bubba/row288 7c01a5ad7 (static) | E1-E3 and E6 fixed; E4 superseded by mos-1520 (notice after push, test must flip); nothing run
 2026-10-23T19:36Z | Webby (Opus 5.5, reviewer seat) | row 288 follow-up on bubba/row288 cef503cb9 (static) | E4 closed per mos-1520; N1 served stays old on notice-store failure; E5 waits on 284; nothing run
 2026-10-23T19:42Z | Webby (Opus 5.5, reviewer seat) | row 198 r4 round 2 static review of ab6417475 (bushroot-r2-sha-webby, mos-1522) | CHANGES: B1 four more tests break on the canSend gate (relaunch:48, shell:215, sso-browser:91, board-refused:122), B2 discord :43/:50 connector text order, B3 conv-no-reply; N1-N3; nothing run
+2026-10-23T19:44Z | Webby (Opus 5.5, reviewer seat) | row 217 early-look follow-up on b6c3129d2 (beakley-0010) | E1 E2 E6 E7 closed statically, E3 premise corrected (organization.set exists), E5 recorded, E4 tests ahead of the fix, M5 anchor safe; F1 minor; nothing run