chore(agent-memory): upgrade 4.42.0 → 4.42.1 — declare a fact consulted to make a decision; check for declaration gaps before archiving one

.agent/schema.md

@@ -233,7 +233,7 @@ Bullet list (if any).
 What the next agent needs to know.
 
 ## Memory References
-- Referenced:  <continuity fact ids this session relied on / reinforced>
+- Referenced:  <fact ids this session relied on (shaped a decision) or reinforced — a read that shaped nothing is not a use>
 - Created:     <new fact ids added this session (born tier: working)>
 - Reactivated: <fact ids pulled back from the archive>
 - Closed:      <thread ids completed this session — the close record is the completion event (or list them under Referenced with a (closed) note)>

.agent/version.md

@@ -4,7 +4,7 @@
 > Mode B can detect drift and upgrade in place (see the tool's `UPGRADE.md`).
 > `version` gates the upgrade ladder — don't hand-edit it unless you mean to.
 
-- **version:**       4.42.0
+- **version:**       4.42.1
 - **enabled_with:**  4.38.0
-- **last_upgraded:** 2026-09-18
+- **last_upgraded:** 2026-10-04
 - **mode:**          A

DECAY.md

@@ -81,7 +81,9 @@ Every session log carries a `## Memory References` section:
 - Reactivated: drizzle-over-prisma
 ```
 
-- **Referenced** — ids the session relied on or reinforced.
+- **Referenced** — ids the session relied on or reinforced. *Relied on* means consulted to make
+  a decision: the session would have decided differently without the fact (v4.42.1). Reading a
+  fact that shaped no decision is not a use.
 - **Created** — new facts added this session (born `tier: working`).
 - **Reactivated** — ids pulled back from the archive.
 - **Closed** — threads completed this session (v4.41.0): list them under `Referenced` with a
@@ -91,7 +93,9 @@ Every session log carries a `## Memory References` section:
 The pre-commit `memory-lint` advisory `[undeclared-reference]` (v4.41.0) warns when a change
 edits a fact's body without declaring the id in a session log staged with it — the diff is the
 only place that omission is visible, because the footers and this log then agree with each other
-while both are wrong.
+while both are wrong. A *consultation* leaves no diff at all, so no check can see an undeclared one
+(v4.42.1): the agent declares it when it writes the log, and the review's subject read before
+archiving a fact as faded (`REVIEW.md` step 6, *declaration gaps*) is the backstop.
 
 So, for any id:
 - `uses` = number of session logs whose `## Memory References` name it.

REVIEW.md

@@ -136,6 +136,22 @@ stalled thread as `[thread-stale]`, so the condition cannot hide.
    Either way, then confirm **no id lives in both `continuity.md` and the archive** (a fact exists in
    exactly one place). Record the result in the summary. (Superseded facts are exempt — they archive
    on truth-state, not recency.)
+
+   **Declaration gaps (facts, v4.42.1)** — the fact-level twin of step 5's thread rule. Both checks
+   above count only *declared* uses, and a fact consulted to make a decision leaves no diff, so for
+   **each** fact archived as *faded*, read for its **subject** — the code, rule or contract it
+   records, not its id. Search the `archive_window` session logs for the subject's distinctive terms
+   (backticked identifiers, words from its bold title), skipping `## Memory Review` and
+   `## Memory References` blocks, and read the sessions that hit. If one **exercised** the subject —
+   changed, tested, applied or decided by it — without declaring the id, the fade is a declaration
+   gap, not disuse: move the fact back as above, name it under *this* review's `## Memory
+   References` (re-affirmed, citing the session that relied on it) so its count resets, and note
+   the reversal in the `## Memory Review` block. A mention is not an exercise: prose that names the
+   subject or the id — a prior review summary, a decay note, a plan never acted on — is not evidence
+   of use (the `ot-review-step6-prose` livelock). The read is judgment and never counts on its own;
+   only the declaration it prompts does. (Field report: mercury-composable and mercury, 2026-10-04
+   — three in-use facts archived in five weeks; this repo's `git-hook-fragment-dispatch` was a
+   fourth.)
 7. **Verify invariants (cadence).** If `sessions_since_last_invariant_check ≥
    verify_invariants_every` (or `last_invariant_check` is unset and that many session
    files exist), raise **one** Open Thread listing every never-decay fact —
@@ -176,7 +192,8 @@ stalled thread as `[thread-stale]`, so the condition cannot hide.
    **re-arms the over-archival guard, and forces a false reactivation** (it will demand you move the
    fact you just archived back). `## Memory References` records only the ids you genuinely
    **relied on / created / reactivated** this session (e.g. a new `(knowledge-harvest)` or
-   invariant-reverify thread you created). The `## Memory Review` block is *not* parsed as references,
+   invariant-reverify thread you created) — and a fact kept after a declaration gap (step 6),
+   which was moved back, not archived. The `## Memory Review` block is *not* parsed as references,
    so archived ids belong there. *(Learned the hard way: a review summary that listed its archived ids
    under `## Memory References` threw 13 spurious `over-archived` ERRORs.)*
    **Inspecting a stalled thread as gate evidence is not a use either** — list a stalled thread

memory/PROTOCOL.md

@@ -80,10 +80,12 @@ on upgrade; fork it under a new name or upstream a genuine fix instead of editin
 - Treat `memory/continuity.md` as working memory and check existing decisions before
   proposing a conflicting change.
 - Note facts, decisions, preferences, and thread changes for session close.
-- Track every fact id referenced, created, reactivated, or closed for the session log's
-  `## Memory References` — an edit or a closure is a use, inspecting alone is not (the pre-commit
-  hook's `[undeclared-reference]` advisory catches a fact edited without a declaration); do not
-  edit `uses`, `last_used`, or `tier` mid-session.
+- Track every fact id relied on, created, reactivated, or closed for the session log's
+  `## Memory References`. A fact is relied on when it shaped a decision — you would have decided
+  differently without it. An edit or a closure is always a use; a read that shaped nothing is not.
+  The pre-commit `[undeclared-reference]` advisory catches an undeclared edit, but nothing can
+  catch an undeclared consultation — declare it when you write the log. Do not edit `uses`,
+  `last_used`, or `tier` mid-session.
 - At a natural seam—milestone, phase shift, or unrelated pivot—persist the session log and
   continuity update before compaction. Context-window utilization is the real pressure
   signal; wall time and perceived vagueness are only proxies. At high utilization, suggest

memory/sessions/2026-10-04-164021.md

@@ -0,0 +1,18 @@
+# Session (2026-10-04T16:40:21.625Z)
+
+**Agent:** Claude Code
+
+Lightweight: Mode B upgrade 4.42.0 → 4.42.1 (PATCH — declare a fact consulted to make a decision; check for
+declaration gaps before archiving one: the protocol now tracks every fact id *relied on* — it shaped a decision,
+you would have decided differently without it; an edit or a closure is always a use, a read that shaped nothing is
+not — and `REVIEW.md` step 6 gains *declaration gaps (facts)*: before archiving a fact as faded, read the window's
+logs for its subject and keep a fact whose subject was exercised without a declaration; from the mercury-composable
+/ mercury field report). Reconcile re-copied DECAY.md, REVIEW.md and .agent/schema.md (each stock 4.42.0 before —
+staleness only). Protocol — re-copied (byte-identical to the 4.42.0 template). The governance pair is not adopted —
+listed as two optional notes (reference only). Adapters re-synced; stamped 4.42.1; every re-copied file
+byte-identical to its MANIFEST source. Ran in a throwaway worktree of origin/main (`--forge github`). memory-lint:
+0 errors, 1 warning(s). Upstream release: https://github.com/Accenture/mercury-go/releases/tag/v4.42.1 .
+
+## Memory References
+
+(none)