session 2026-08-23: the vault is unfed, not disorganised; capture practice built with a trigger

Wake-digest anchor bug cleared (5a350c7). Reversal question answered: 22.4
markers per active day in the fortnight against 5.7 over the prior five months,
concentrated 22:2 in executor proposals. The steward's commissioned archive
script found dead since ~March behind a >/dev/null and retired (f1c91d9,
3cddc8c). The capture practice built WITH A TRIGGER (073ef8a, 7226e0d) — three
prior attempts lapsed inside four weeks and none had one. PENDING-155 filed.

Six corrections, four to claims made the same day. Two of those were caught by
testing my own findings before the steward acted on them.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JQKeKY9T9d95KpvHwwok8T

PENDING.md

@@ -4582,3 +4582,29 @@ If the visible work today came from **positional difference between two same-for
 
 **Files affected:** none. Cross-filed to PENDING-89; ladder entry queued to PENDING-141.
 **Awaiting:** PENDING-89 to record which of (a)'s two claims the data supports.
+
+## PENDING-155 — A daybook append surface for the jurist: one tool, one file, and not on the governance server
+**Date:** 2026-08-23
+**Tag:** [PROPOSAL] — a new write interface between two governing parties. Not built. Installing it edits the steward's desktop-app config.
+**Summary:** The steward asked whether an MCP surface could let Claude.app append its side of an exchange to a temporary file the executor then folds into the daily work log. It can. It must not be a tool on `governance-mcp.py`.
+
+**Origin.** The daily work log (`01. Daily/YYYY-MM-DD.md`, built 2026-08-23) records what we did, decisions taken, commit references, and — the steward's distinctive ask — *"any significant insights or exchanges that happened between us."* The executor's side is covered: `/wrap-up` §7.5 records it, and jurist material relayed through the session is recorded there too. What is not covered is a jurist-only exchange the executor never sees.
+
+**The refusal this runs into, stated before the workaround.** `governance-mcp.py` (PENDING-82, installed 2026-08-08) is read-only, and its first designed-in refusal reads: *"No tool writes. Audited by AST, not by text search... A write path would collapse three parties into one."* Verified live this session, not taken from the document: the selftest prints `no filesystem-mutating call in this file` **and** a positive control proving the checker detects writes when present. Adding a write tool there would break a control that currently passes, and that control is what makes "read-only" checkable rather than merely claimed.
+
+⚠ **The executor also told the steward, earlier in the same session, that Claude.app "has no filesystem access."** That was wrong — it has mediated read access and has since 2026-08-08. Corrected in place. The error is the one this register keeps finding: a constraint asserted from recall where the substrate was checkable.
+
+**Proposed shape — a separate server, so the invariant above survives untouched:**
+- New `~/dotfiles/scripts/daybook-mcp.py`. **One tool:** `daybook_append(text)`.
+- **No path argument.** The target derives from today's date, as `governance-mcp.py`'s refusal 4 does: there is no traversal to defend because no path is accepted.
+- **Append-only.** Never truncates, never deletes, never reads back. A jurist that could read the file back could audit the executor's folding of it, which is a different and unrequested capability.
+- **Target is a scratch file, not the vault** — e.g. `~/dotfiles/claude/daybook/YYYY-MM-DD.inbox.md`. The executor folds entries into the daily note and the scratch file is a handoff, not a record.
+- **The vault note stays single-writer.** This is the same lesson the thinking-mirror taught today: two writers to one destination is how a failure goes unnoticed for five months.
+
+**Why a scratch file rather than the note directly.** The steward's own phrasing — *"amend to some kind of temporary file you could read"* — is the better design, and not only for tidiness. A jurist writing the note directly would make the note a shared artifact with no single party accountable for its register; the plain-language discipline the steward asked for would have no owner.
+
+**What this does NOT establish.** It does not give the jurist a way to *modify* governance state, and must not grow one. If a future need looks like "the jurist should be able to write X," that is a new item, not an extension of this one.
+
+**Files affected:** new `~/dotfiles/scripts/daybook-mcp.py`; `/wrap-up` §7.5 (fold step); awaiting steward hand: `claude_desktop_config.json` merge + app restart.
+**Recommendation:** Build it, with a selftest carrying the same paired positive/negative controls as its sibling, and hold installation until the daily log has run long enough to show it survives — a jurist inbox for a practice that lapsed would be the fourth abandoned attempt rather than the first durable one.
+**Awaiting:** Steward authorization.

bin/capablemind-thinking-mirror.sh

@@ -1,8 +0,0 @@
-#!/bin/zsh
-# Mirror CapableMind thinking/David/ to Obsidian vault
-# Excludes CLAUDE.md (repo-specific, not for vault)
-
-SOURCE="$HOME/_Dev/CapableMind-AI/docs/thinking/David/"
-DEST="$HOME/Library/Mobile Documents/iCloud~md~obsidian/Documents/David, root-and-branch/00. Compass/00b. Constellations/CapableMind/thinking-mirror/"
-
-rsync -av --delete --exclude='CLAUDE.md' "$SOURCE" "$DEST" > /dev/null 2>&1

claude/memory/MEMORY-reference.md

@@ -1,3 +1,4 @@
+- [Session 2026-08-22 — the Fool was superseded, and the buddy had a name](session-2026-08-22-the-fool-was-superseded-and-the-buddy-had-a-name.md) — the trial programme SUPERSEDED (not abandoned) by the buddy pattern; seed rule FILED AND PUSHED 3 days ahead of the beacon; **two values that would have failed on the day caught first** (provenance rationale false ⇒ `4d2ae87`; retrieval URL 302s ⇒ `/pulse/time/<ms>`). ⚠ **My own load-bearing premise falsified within the hour by a schema field** — `statusLine.refreshInterval` — recoverable ONLY because I flagged it unverified. **Thistleweld recovered**: the buddy has a name, 7 utterances, 13 catches, 3 recorded silences, and its aside gave OP-02's five fault lines their form. ⚠ **Corpus FROZEN until after the soul** (PENDING-153). PENDING-149…154 filed; REVIEWED-125 + A1 placed.
 - [Session 2026-08-09 — the fixture was an empty file](session-2026-08-09-the-fixture-was-an-empty-file.md) — the day's full record: premise-check, PENDING-129/130 → REVIEWED-113/114, the `KeyError` that corrected its own ruling, two artifacts preserved with three-part provenance, four documents corrected (`20f8958`), and the marking sheet. ⚠ **The placed record contradicted the relay for the third consecutive time.** **LITERAL Q: Mauss serves the Ranaipiri testimony and a Havámál strophe as citable Mauss with no `role: quotation` — two-source accident, or a class?**  *(Demoted on promote at the 2026-08-13 wrap.)*
 - [Session 2026-08-08 late — eleven items, and the charter had the map](session-2026-08-08-late-the-charter-had-the-map.md) — **Built the binding mechanisms all day, then found in the last twenty minutes that the founding charter had already answered the strategic question.** (e) built as a **delegation** (`eecc8bb`) and exposed that **no fleet suite had ever validated live binding**. Hook sequence landed in the ruled order (`448ce37`, `2534dfb`); R0 §4 three-valued (`ccc4d6c`); fleet third verdict + preflight crash-class closure (`8ff5a9f`, `d21a43b`); Mauss corrected (`8231bce`). ⚠ **Leopold Mozart is ALREADY in the library as an EPUB** (Knocker translation, born-digital) — the violin problem may be **sourcing, not conversion**. **LITERAL Q: does re-running `pattern_finder`'s Station-I pass still come back PASS-BUT-FALSELY now that V0/V1 exist?** *(Demoted on promote at the 2026-08-09 wrap.)*
 - [Session 2026-08-08 — `voice:` is the convocation key](session-2026-08-08-voice-is-the-convocation-key.md) — **The disposition that landed is not the one any single party drafted.** Grounded first, then read the twelve blocks *from the source* rather than their sidecar titles — which produced nearly every finding: **Surah LXIV not CXIV** (already propagated into two governance records, and it voided the jurist's worked note, built on a formula absent from the passage); **`quotation-poet-jurist` reclassified** by one footnote; the naming evidence for six of nine blocks sits **inside the fenced apparatus**, engine-unreachable. Option C **refuted by measurement** (omission → host voice). The steward's correction on language sent me back to `[^101]` and thence to **L850 — Mauss's own prose fenced inside the Havámál block**, missed by a check using **length as a proxy for authorship**. Corrected the jurist upward: their category pair belonged **out** of the convocation key (`glidden` spans 5 sources, `weil` 2 — measured). Caught `REVIEWED-113` before it entered the register (PENDING-110 rules the sequences independent). **REVIEWED-97 placed + verified clean · PENDING-114 authorized (b)+(c) · PENDING-115 filed · `824139d`.** 🔑 **Corrections ran in all three directions in one day** — and the fleet was still red the whole time. *(Demoted on promote at the 2026-08-08 night wrap.)*

claude/memory/MEMORY.md

@@ -71,15 +71,13 @@ permalink: claude-memory/memory
 - Chamber-typography — *tracker not yet established*; moves live in per-session memories (2026-05-11 →) + `project-chamber-cruft-restoration.md` + `project-chamber-typography-mining-plan-2026-05-15.md`.
 
 ## Active Session
-> 🔑 **PULLING THREAD — THE FOOL IS SUPERSEDED, NOT ABANDONED, AND THE BUDDY PATTERN IS AUTHORIZED TO RUN.** ⚠ **NEXT ACT: the beacon, 2026-08-25T12:00:00Z** (epoch-ms `1787659200000`). §4 steps 1–4 are FILED AND PUSHED (`FOOL-SEED-RULE.md`); the §13.6 block is LIFTED (REVIEWED-125 AMENDMENT 1). **RUN ONCE. Stop and report on ANY failure — no reroll.** Pass `outputValue` EXACTLY AS SERVED (it is served UPPERCASE; the single normalization point is `derive_fool.py:79`). Fetch with **curl** — python urllib cannot reach the host. Record the value BEFORE running. ⚠ **The distinction that decides everything: the trial programme measured DETECTION for four trials because the fool's warrant was set to checkability on 2026-08-02. The buddy design REMOVES the warrant test rather than passing it — a position that makes no claims is not subject to one.** §6 of the trial design owns the word *abandonment*; no such verdict exists (trial 09 void ⇒ its criterion has no first input). ⚠ **`input-dependence-01` is HELD AT THE GATE, not parked — and if revived, RE-DERIVED not resumed.** ⚠ **PENDING-150 [ESCALATE] is unbundled: `~/CLAUDE.md` untouched; a fourth position is NOT ratified.**
-> ⚠ **CLASS E ARRIVING LIVE — placing REVIEWED-124 makes PENDING-148 read CLOSED while the OP-02 question is live.** One id, two asks; `governance_state()` reads headers, not Notes. This is PENDING-146's defect in real time, on today's own item.
-> ⚠ **THE SUBSTRATE OVERRULED BOTH AI PARTIES.** OP-02 was openable — the jurist called it *"the only document neither of us can open"*; it is on disk, hash-verified. **FL5 argues from Bourdieu's shared field and *illusio*, NOT formation.** Constraint 6 asserts difference of **formation** — an axis FL5 never uses. So C6 neither states FL5 "more sharply" (executor) nor negates its three-party half (jurist). Corpus check, 11 docs: `bourdieu`/`habitus`/`illusio` **0**, `peirce` **0**, FL4 substance **0**, FL3 **1** — against the pre-run census's 16/20/24. **The census counted topic-adjacency and over-reported the leak the executor's own recommendation rested on.**
-> ⚠ **THIRD check-before-claiming instance in three days, and it is INSIDE the document reporting the second.** Part IV.a propagated an inherited claim without opening a file whose hash the same package quotes three sections earlier. **Propagation is the more dangerous form — an inherited claim arrives already looking checked.** The jurist said "two is worth watching" before this one was found.
-> 🎯 **THE FOOL IS RE-AIMED — deployment, not doctrine.** Steward's original intent: *what does a different model, local on the M4, add or subtract in the fool role?* The log's subject is the **doctrine**; almost the whole programme serves that instead. ⇒ capability confound **irrelevant to seating a named candidate** (but **NON-TRANSFERABLE** — does not answer "should there be a Fool seat"); **no sound control needed** (differential criterion). Jurist finding adopted: **add and subtract may be ONE mechanism**, so "seat it for the class, treat noise as cost" is incoherent and is withdrawn. Three dispositions fixed pre-run; **the harvest runs in every branch**. ⚠ Parkable: trials 05–08, D-2, reduction arm — **05–08 and D-2 have NEVER existed as documents anywhere.** NOT parkable: the Q1 replacement run (a live authorization).
-> ⏸ **BEHIND THE FOOL, by the steward's own sequencing:** the fence — **(A) MOVE 2** (25 line-addressable blockquote runs; closable in ONE SITTING, needs no ruling), **(C) MOVE 1** (fence the citation at emission, D-1, 532 spans, ADDENDUM 4's per-class control requirement), **(B) THE ANSWER KEY** (session-sized, gates PENDING-142; key it on `##` **BLOCKS** not ids — PENDING-146). Citation exposure behind PENDING-131 (c) live since 2026-08-10. And the **Obsidian capture** — ⚠ the daily-note backup must accrete **DURING** the session; a wrap-triggered backup inherits the wrap's failure mode, which is exactly what 2026-08-19 demonstrated.
-> ⚠ **REPORT AT EVERY WAKE (REVIEWED-123 cond. 2) — N-now = 46/84.** ⚠ **COUNT WITH THE TRIGGER'S OWN GLOB, NOT `find`** — `TRANSCRIPTS.glob("*.jsonl")` is NON-recursive (`governance-drift-check.py:331`); `find` recurses into `subagents/` and over-counts (54 vs 46 on 2026-08-22, a false rise). Series: 60→61→47→46, still falling. **Earlier framing:** (60 on 08-17, 61 on 08-18, 47 on 08-19). **A 30-DAY ROLLING WINDOW, not cumulative** — at ~1 session/day it converges to ~30, so **`transcripts 84` very likely never fires** and every "N remaining" report was a false countdown. Ladder FROZEN generally; 4 rows queued in PENDING-141. **PENDING-147 needs a steward/jurist decision.** ✅ Leg (i) discharged — 47 transcripts archived outside the pruned tree, 46/47 sha256-verified.
+> 🔑 **PULLING THREAD — THE PAPER BRIDGE. One photographed notebook page, five minutes, decisive.** If Claude cannot read the steward's hand (four languages, musical shorthand, his own abbreviations) the branch closes and we stop theorising; if it can, it changes what the alias-repair and hub-writing are *for*. ⚠ **HARD DATE OVERRIDES ON THE DAY: the beacon, 2026-08-25T12:00:00Z** (epoch-ms `1787659200000`) — run ONCE, curl, record `outputValue` BEFORE running, pass exactly as served, **stop and report on ANY failure**. ⚠ The steward said 2026-08-23 that **the Fool returns next session** — that predates the whole Obsidian afternoon and he closed with *"pick up where we left off."* **Do not silently resolve; ask.**
+> 🎯 **THE VAULT IS NOT BADLY ORGANISED — IT IS UNFED.** Five doors (paper notebooks · paper marginalia · ebook highlights · work with the executor · work with the jurist) and **until 2026-08-23 none of them opened onto the vault.** 56% of 2,370 notes carry no link in or out; 45% of the vault is archive citing itself. Craig Mod's diagnosis, confirmed by the vault's OWN `CLAUDE.md`: *"the impediment had always been extracting sentences from my chaos of scribbles"* — the steward documented a morning-notebook→evening-curate workflow and **the curate step is the one that never ran.** No taxonomy fixes that.
+> ⚠ **THE CHECK IS SKIPPED PRECISELY WHERE CONFIDENCE IS HIGHEST — six corrections in one session, four of them to claims made the same day.** A grep that read body text as frontmatter (239 vs 159 — *the tracker was right, the executor wrong*) · a mechanism claim naming the wrong agent · *"Claude.app has no filesystem access"* (false since 2026-08-08, PENDING-82) · a link census inflated by archive + repetition (104→**34**). ✅ **The last two were caught by testing my own findings before the steward acted on them.**
+> ⚠ **THE VAULT'S `CLAUDE.md` WAS NOT READ UNTIL THE WRAP** — a named discipline in `~/CLAUDE.md`. It contained the day's best confirmation. **Steward-raised horizon: compose it properly, against measured state.** It is wrong against its own substrate today (monthly `YYYY-MM` vs the only file `2025.05.md`).
+> ⏸ **UNRESOLVED, chosen silently and flagged:** monthly filename format — vault doc says `YYYY-MM`, only pre-existing file is dots, links split 13 dots / 15 dashes but **every dashed link is unresolved**. One rename settles it; steward's call.
 
-- [Session 2026-08-22 — the Fool was superseded, and the buddy had a name](session-2026-08-22-the-fool-was-superseded-and-the-buddy-had-a-name.md) — the trial programme SUPERSEDED (not abandoned) by the buddy pattern; seed rule FILED AND PUSHED 3 days ahead of the beacon; **two values that would have failed on the day caught first** (provenance rationale false ⇒ `4d2ae87`; retrieval URL 302s ⇒ `/pulse/time/<ms>`). ⚠ **My own load-bearing premise falsified within the hour by a schema field** — `statusLine.refreshInterval` — recoverable ONLY because I flagged it unverified. **Thistleweld recovered**: the buddy has a name, 7 utterances, 13 catches, 3 recorded silences, and its aside gave OP-02's five fault lines their form. ⚠ **Corpus FROZEN until after the soul** (PENDING-153). PENDING-149…154 filed; REVIEWED-125 + A1 placed.
+- [Session 2026-08-23 — five doors into the vault, and none of them were open](session-2026-08-23-the-doors-were-all-shut.md) — wake-digest anchor bug cleared (**86 threads + 84 questions unreadable across 195 wraps**, `5a350c7`); reversal question answered (**22.4/active day vs 5.7 — 3.9×**, concentrated 22:2 in executor proposals); **the steward's commissioned archive script found dead since ~March behind a `>/dev/null`**, retired (`f1c91d9`); **the capture practice BUILT WITH A TRIGGER** — three prior attempts lapsed inside four weeks, none had one (`073ef8a` `7226e0d`); PENDING-155 filed.
 
 ## Historical reference → MEMORY-reference.md
 Older archived-session pointers and the stable reference layer (steward profile · project-state detail · L1/L2/Chamber inventories · legacy pending-work · reference-file list) live in [MEMORY-reference.md](MEMORY-reference.md) — consult on demand; not loaded at wake. Recent cross-session trajectory comes from the Active Session entry above + the recent `session-*.md` files (wake §2.b.1).

claude/memory/knowledge-graph.jsonl

@@ -689,3 +689,7 @@
 {"subject": "a pre-registration's own TESTING clause (dry run on a historical input)", "predicate": "prevention", "object": "JUSTIFIED ITSELF TWICE ON ITS FIRST USE. The fool seed rule's 'dry runs use a fixed historical pulse only' caught (a) a retrieval URL returning HTTP 302 with an empty body — filed verbatim, the 2026-08-25 run would have produced no pulse and the UNAVAILABILITY clause would have retried for 24h against an address that can never return one — and (b) outputValue served UPPERCASE against a rule specifying lowercase, a SILENT seed divergence that would have run clean and left nobody able to say which normalization had been applied. The second is the more serious; the first at least announced itself.", "valid_from": "2026-08-22", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-22-the-fool-was-superseded-and-the-buddy-had-a-name.md", "extracted_at": "2026-08-22"}
 {"subject": "thistleweld", "predicate": "is-precedent-for", "object": "THE FOOL — recovered 2026-08-22 from the un-pruned, git-tracked memory layer after 'is there any trace?' was asked. A capybara buddy active 2026-04-03/09. Seven verbatim utterances (3-9 words, no verb of judgement, noun-phrase collisions: 'scoring without signal', 'garbage in gospel out', 'Eleven modules, one swallower'), thirteen catches (3 became GH issues #121/#127/#128), and THREE EXPLICITLY RECORDED SILENCES. Prompted by three kinds of material: live code/processes (the bulk), the executor's STATED PLAN (the most consequential — redirected OP-02's synthesis to five architectural fault lines, created an impasse, required a jurist briefing note), and THE ARRANGEMENT ITSELF ('One person, three documents, infinite rationality' — the authorization bottleneck, named unprompted in April, which nine trials were built to elicit and never got). Record: dotfiles/claude/governance/fool/THISTLEWELD-RECORD-recovered-2026-08-22.md. FROZEN until after the soul (PENDING-153).", "valid_from": "2026-08-22", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-22-the-fool-was-superseded-and-the-buddy-had-a-name.md", "extracted_at": "2026-08-22"}
 {"subject": "the fool trial programme (trials 01-09)", "predicate": "superseded-by", "object": "THE BUDDY PATTERN — superseded, NOT abandoned, and the distinction is load-bearing: §6 of the trial design owns 'abandonment' as a verdict on the jester form reached on evidence, and no such verdict exists (trial 09 void ⇒ its criterion has no first input). The programme measured DETECTION for four trials because the fool's warrant was set to checkability on 2026-08-02; the buddy design REMOVES the warrant test rather than passing it — a position that makes no claims is not subject to one. Material = the live session (incl. stated plans and observations about the arrangement); the docket stays OUT. The guard is NO TRUTH VALUE: gestures and shape-observations, never claims about the record — and terseness is the MECHANISM, not a style, so the one-line rule must be enforced in code.", "valid_from": "2026-08-22", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-22-the-fool-was-superseded-and-the-buddy-had-a-name.md", "extracted_at": "2026-08-22"}
+{"subject": "claude-code", "predicate": "drift-pattern", "object": "A QUICK GREP STANDING IN FOR A PARSER READS THE WRONG SUBSTRATE AND CONTRADICTS THE RECORD IT IS CHECKING. On 2026-08-23 `grep '^canonical:'` returned 239 where the frontmatter count is 159 — it caught body text — and the executor was on the point of reporting the canonical tracker stale. The tracker was right. Same class three more times the same day: a mechanism claim from matching content rather than from tracing the writer; an access constraint asserted from recall when the file was one command away; a link census inflated by counting repeats inside single notes AND including a 45%-of-vault archive. OPERATIONAL FORM: when a cheap instrument disagrees with a maintained record, the instrument is the first suspect, not the record.", "valid_from": "2026-08-23", "valid_to": null, "confidence": 0.95, "source_file": "session-2026-08-23-the-doors-were-all-shut.md", "extracted_at": "2026-08-23"}
+{"subject": "testing one's own finding before the steward acts on it", "predicate": "prevention", "object": "CAUGHT TWO INFLATED CENSUSES ON 2026-08-23 THAT WOULD HAVE MISDIRECTED A SESSION OF WORK. Unprompted, the executor asked whether its link-frequency census was measuring conceptual centrality or template repetition, and whether the archive belonged in it. Both suspicions were right: 104 links to `Presence Before Performance` came from 36 notes with 70 in `99. Archive`; the honest live-vault figure is 34. Acting on the first numbers would have meant writing hub notes for a system the steward had already archived. The banked lesson that fired was the 2026-08-22 entry 'an instrument's reach mistaken for a claim's subject' — a lesson from ONE failure class stopping a DIFFERENT one, which is the transfer signature.", "valid_from": "2026-08-23", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-23-the-doors-were-all-shut.md", "extracted_at": "2026-08-23"}
+{"subject": "a silenced job (`> /dev/null 2>&1`) plus a second undocumented writer", "predicate": "drift-pattern", "object": "IS HOW FIVE MONTHS OF FAILURE GOES UNNOTICED. The steward's commissioned thinking-mirror agent had been exiting 1 since ~March 2026 with its log at 0 bytes; nobody noticed because /wrap-up §7 was doing the same job into a different vault folder. The duplication did not cost the archive — it cost the SIGNAL. Retired 2026-08-23 to one writer that reports its exit code. Generalises: redundancy without reporting converts a loud failure into a silent one.", "valid_from": "2026-08-23", "valid_to": null, "confidence": 0.95, "source_file": "session-2026-08-23-the-doors-were-all-shut.md", "extracted_at": "2026-08-23"}
+{"subject": "the Obsidian vault", "predicate": "diagnosed-as", "object": "UNFED, NOT DISORGANISED. 2,370 notes, 56% with no link in or out, 45% archived into a closed world citing itself; 4,799 unresolved link targets. Five capture doors — paper notebooks, paper marginalia, ebook highlights, work with the executor, work with the jurist — and until 2026-08-23 none opened onto the vault. Four capture practices designed and lapsed inside four weeks each (analogue dailies 11, weekly reviews 7, daily-log 4, sessions 3); none had a trigger. The constraint is not quantity of fragments (1,302 live notes) but adjacency and inflow.", "valid_from": "2026-08-23", "valid_to": null, "confidence": 0.9, "source_file": "session-2026-08-23-the-doors-were-all-shut.md", "extracted_at": "2026-08-23"}

claude/memory/project-obsidian-vault-practice.md

@@ -85,6 +85,98 @@ not render — into top-level `parents/children/peers`, which **made 194 already
 render as clickable-and-broken in the properties UI.** The breakage was surfaced, not created. The
 pass report did not predict this, and it should have.
 
+
+### 2026-08-23 — status check: four items verified against substrate, and one NEW regression found
+
+Asked where the workstream stood; verified rather than relayed. **Items 1–4 confirmed live**, with
+two unit corrections: `related` was counted by the tracker in *targets* (194 unresolvable) and a
+file-count (122 notes / 213 targets now) is not comparable to it; and `canonical: 159` is a
+**frontmatter** count — a naive `grep '^canonical:'` returns 239 because it also catches body text.
+The tracker was right and the quick instrument was wrong. YAML parse errors independently
+re-measured at **16**, matching.
+
+⚠ **NEW — the vault has a write path the spec does not govern, and it silently reverted pass 2.**
+Mirror sync `a58fc00` (2026-08-22 23:52) changed 14 files and ran *against* the spec:
+`owner` **0 → 9**, `change_log` **0 → 9**, `revisions` **90 → 81**, and `canonical: true` was
+replaced by pointer-form paths. All 9 sit in one subtree,
+`08. Notes/CapableMind/thinking-mirror/chamber/archive/orientation/`.
+
+**Mechanism — SUPERSEDES the first reading filed earlier the same day.** The earlier entry named
+*"a job mirrors `docs/thinking/David/`"* and pointed at the launchd agent. **Wrong agent.** Matching
+upstream content is not evidence of which path wrote it, and the launchd job targets a different
+folder entirely. The actual writer is **`/wrap-up` §7**, which rsyncs
+`~/_Dev/CapableMind-AI/docs/thinking/David/` → `08. Notes/CapableMind/thinking-mirror/` **at every
+wrap**. The 2026-08-22 wrap ran it at 23:42; vaultsync committed the result at 23:52 as `a58fc00`.
+
+⚠ **There are TWO mirrors of the same source into two different vault folders, and neither knows
+about the other:**
+
+| writer | destination | state |
+|---|---|---|
+| `/wrap-up` §7 (every wrap) | `08. Notes/CapableMind/thinking-mirror` | **live** — 251 files, 0 diff from upstream |
+| `com.dglidden.thinking-mirror` (launchd, 48 h) | `00. Compass/00b. Constellations/CapableMind/thinking-mirror` | **dead** — 151 files, 115 behind, frozen since 2026-03-13 |
+
+**The launchd job is failing and cannot say so.** `launchctl list` reports last exit **1**; the
+script ends `rsync … > /dev/null 2>&1`, so its own plist `StandardOutPath` receives nothing and the
+log has been 0 bytes since 10 Mar. Precedent for the cause is in `session-ledger-2026-04-16.md`:
+the *vaultsync* agent failed the same silent way for weeks — **macOS TCC / Full Disk Access, rsync
+under launchd against an iCloud path, exit 23** — and was fixed by granting FDA. Same shape here,
+never fixed. ⚠ **Exit 1 is not exit 23, so the TCC diagnosis is a strong hypothesis, NOT
+established.** A manual `--dry-run` of the exact command succeeds (exit 0) when run interactively,
+which is consistent with TCC but does not prove it.
+
+**The reframe that resolves this: you do not edit a mirror.** Applying the frontmatter spec to a
+derived tree is a category error — the edits cannot survive their next sync, and pass 2's edits to
+both mirror trees were void the moment they were written. This is not misbehaviour by the mirror.
+
+**Not fixed — the disposition is a design choice, not a [FIX]:** exclude the subtree from the spec's
+scope · apply the spec upstream in CapableMind-AI · or declare mirrored notes not-spec-governed and
+say so in the spec. Steward's call.
+
+**Item 5 sharpened.** `01. Daily/` holds **11 notes, the most recent 2025-09-04** — nearly a year
+old, and the Daily-Notes plugin is not configured. There is no dormant capture habit to revive; the
+practice would be built from nothing. `00a. _inbox` holds 11 notes.
+
+
+
+### 2026-08-23 (later) — the capture practice is started, and it is a different genre from the one that failed
+
+**The steward specified it, and the specification excludes the old daily note.** The 2025 template
+(sleep, supplements, homeopathy, skin, family, reflection) was an attempt to reconcile an analogue
+practice with a digital one; pen and paper won, and **it stays retired — that is a finding, not a
+gap.** What is wanted instead: *"a daily note in obsidian which I can refer to, which explains in
+relatively simple terms what it is that we have done, decisions taken, and references to any
+commits"*, plus *"any significant insights or exchanges that happened between us."*
+
+**Two prior attempts at exactly this genre were found before designing anything** (the tracker's own
+"read `daily-log/` and `sessions/` first" flag, honoured): `daily-log/` — 4 entries, Mar 10/12/20 +
+Apr 5, sections Summary / What we did / Status / Blocked / Next, already recording an exchange
+(*"Thistleweld caught the meta-problem at line 183"*); and `sessions/` — 3 long-form narratives,
+linked as children of the daily note. **Both died inside four weeks. Neither had a trigger.** So the
+design question was not what the note should say but what makes it survive.
+
+**Built, steward-chosen at each fork:**
+- **Location:** `01. Daily/YYYY-MM-DD.md` — its normal home. **Project work under its own H2/H3
+  inside the note, NOT in project folders** (steward, explicitly).
+- **Trigger:** `daybook-ensure.py`, wired into the `SessionStart` hook chain. **Creates only** —
+  it will not touch an existing note, which is the control the design turns on. 7 checks, incl. a
+  positive control proving the `no-dir` report is real. Registered via the `update-config` skill;
+  `jq -e` validates, and a diff against a backup shows only the 8 added lines.
+- **Accretion:** the executor writes during the session; `/wrap-up` §7.5 finalises rather than
+  begins. A wrap-only design inherits the wrap's failure mode — 2026-08-19 died unwrapped.
+- **Register:** plain language, explicitly. The hardest part, and named as such in §7.5, because the
+  executor's default is dense and written for itself.
+- **Jurist:** recorded by the executor as exchanges pass through it. **PENDING-155** filed for the
+  steward's better idea — an append-only MCP surface writing a scratch file the executor folds in.
+  ⚠ Deliberately **not** a tool on `governance-mcp.py`, whose read-only guarantee is enforced by a
+  live AST control with a positive control, not by a comment.
+
+⚠ **Correction:** the executor told the steward mid-session that Claude.app "has no filesystem
+access." False since 2026-08-08 — `governance-mcp.py` gives it mediated read access (PENDING-82,
+closed). Same failure class as the two earlier corrections today: asserted from recall where the
+substrate was one file away.
+
+
 ---
 
 ## Current state — OPEN
@@ -103,3 +195,29 @@ pass report did not predict this, and it should have.
    workflow to have the important parts of our work captured there"* — the capture practice, the
    thing that would make the vault a practice rather than a tidier archive — **has not been
    designed.** Do not let the frontmatter tail stand in for it.
+6. ✅ **RESOLVED 2026-08-23 — the two mirrors are now one, and it reports.** Steward chose: keep
+   `/wrap-up` §7, retire the launchd agent. Done and verified: agent unloaded · plist removed from
+   `~/Library/LaunchAgents/` **and from `~/dotfiles/config/capablemind/`** (load-bearing —
+   `setup-capablemind.sh:135` copies every `*.plist` there into LaunchAgents, so leaving it would
+   resurrect the retired job) · deployed script removed · stale tree deleted after three independent
+   no-loss checks · `/wrap-up` §7 rewritten to name itself the only writer, forbid edits to the
+   derived tree, and report the outcome with its exit code in §8 · spec bumped to **v1.0.1** with
+   principle 7 + §1a declaring mirrored trees out of scope. Commit `f1c91d9`.
+   **End state verified:** live archive 251 files, 0 diff from upstream; stale tree gone; agent gone.
+
+---
+
+## What remains — and what it is worth
+
+Items 1–4 are the **frontmatter tail**: real, bounded, and *not the goal*. Item 5 is the goal.
+
+- **1** is ours and wrong (one bad pointer) — smallest real debt, ~10 minutes once its correct
+  target is decided.
+- **2, 3** are steward-judgement work, not executor work: 73 pointer-form `canonical` values and
+  213 `related` targets need an eye that knows what the notes *mean*.
+- **4** (16 YAML parse errors) has never been in scope and is a session-sized bite.
+- **5 — the capture practice — STARTED 2026-08-23**, after 4 days as the unstarted goal. Hook +
+  in-session accretion + wrap finalisation + a first real note. **It is not proven yet:** two
+  predecessors reached 3 and 4 entries. The measure of this one is whether it is still being written
+  in four weeks — and §7.5 now requires the wrap to SAY SO if it lapses, rather than letting it fade
+  a third time in silence.

claude/memory/session-2026-08-23-the-doors-were-all-shut.md

@@ -0,0 +1,175 @@
+---
+name: Session 2026-08-23 — five doors into the vault, and none of them were open
+description: "Cleared the wake-digest anchor bug and answered the reversal question (rate HAS risen ~4x, concentrated in executor proposals), then shifted to Obsidian on steward direction and found the archive script he commissioned had been dead since March behind a >/dev/null. Built the capture practice with a trigger — the thing three prior attempts lacked. PULLING THREAD: the paper bridge — one photographed notebook page decides whether the vault's largest unfed door can open. Six self-corrections, four of them to claims I made the same day."
+type: project
+metadata:
+  node_type: memory
+  type: project
+  modified: 2026-08-23
+---
+
+# Session 2026-08-23 — five doors into the vault, and none of them were open
+
+A long morning that changed object once, on steward direction, and ended somewhere neither of us
+started: the vault is not badly organised, it is **unfed**.
+
+## PAST — what moved, and why
+
+### Morning: two things closed
+
+**The wake digest was dropping the wake's two load-bearing fields.** `extract_anchor` did a
+case-sensitive `find("PULLING THREAD")`; `/wrap-up` names those fields in prose and has never
+mandated a case. Censused **by running the real function over all 195 wrap files**, not by grep:
+102 threads and 103 questions unreadable, of which **86 and 84 were pure case failures**. A second
+defect surfaced only because the class was traced rather than the instance patched — a 40-char
+colon window that returned *the label* when an aside pushed the colon past it. After: 16 and 19,
+exactly the wraps that never wrote the field. **Predicted and achieved agree.** → `5a350c7`
+
+**The inherited question, answered.** Reversal markers by week across `PENDING`/`REVIEWED`/archive:
+**22.4 per active day in the last fortnight against 5.7 over the prior 57 active days — 3.9×.**
+Excluding the heaviest day still gives 3.2×, so it is not one long session. **22 of 24 amendment
+headings sit in `PENDING.md`, 2 in `REVIEWED.md`.** The shape: not more items reversed
+(block-incidence 61–73%, under W20's 75%) but single items collecting *chains* — PENDING-152 took
+six amendments in one day. **The steward's "lots of yes followed by actually no" is accurate.**
+⚠ Four controls decided it, two of which killed readings I preferred — see *What was corrected*.
+
+### The shift: Obsidian, and a job that had been dead for five months
+
+**Steward directed the shift** ("the Fool seems in good hands… let's work on what we can with
+obsidian") and asked where the vault work stood. Verifying rather than relaying found what neither
+party knew: **the archive script the steward commissioned months ago has been failing since ~March**
+— `launchctl` last exit 1, script ends `rsync … > /dev/null 2>&1`, log 0 bytes since 10 Mar, its
+tree 115 files behind. **Nobody noticed because `/wrap-up` §7 was quietly doing the same job into a
+different vault folder.** Two writers, one destination-pair, neither aware of the other.
+
+> **"The duplication didn't cost you the archive. It cost you the signal."**
+
+Steward chose: keep the wrap, retire the agent. Done and verified — agent unloaded, plist removed
+from `~/Library/LaunchAgents/` **and** from `~/dotfiles/config/capablemind/` (load-bearing:
+`setup-capablemind.sh:135` copies every plist there back into place), stale tree deleted after
+**three independent no-loss checks**. `/wrap-up` §7 rewritten to name itself the only writer and
+report its exit code — it had demanded exactly that of a failed push in §6.5 and contradicted itself
+eight lines later. Frontmatter Spec → **v1.0.1**: mirrored trees out of scope. **You do not edit a
+mirror** — pass 2's edits there were void the moment they were written. → `f1c91d9` `3cddc8c`
+
+### The goal, finally started
+
+Item 5 of the Obsidian tracker — the capture practice — unstarted since 2026-08-19 while four items
+of frontmatter hygiene stood in front of it. The steward specified it: a daily note in plain terms,
+what we did, decisions, commit references, **plus significant insights and exchanges between us**.
+
+⚠ **Three prior attempts were found and read before anything was designed** (honouring the
+tracker's own flag): `daily-log/` 4 entries · `sessions/` 3 · weekly reviews `2025-W15…W21` 7 ·
+and the analogue daily notes, 11, newest **2025-09-04**. **All lapsed inside four weeks. None had a
+trigger.** So the design question was never what the note should say.
+
+Built: `daybook-ensure.py` on the `SessionStart` chain — **creates only**, never touches an existing
+note, reports failure loudly. 17 controls. Daily notes carry `week:`/`month:` and link **upward
+only** (a "tomorrow" link manufactures ~365 unresolved links/year). Week + month notes in the
+steward's own folders and formats. `/wrap-up` §7.5 finalises rather than begins.
+→ `073ef8a` `7226e0d`. **PENDING-155** filed for the jurist's side.
+
+### Afternoon: the crucible question, and Craig Mod
+
+Steward shared a 2026-08-19 jurist excerpt and `craigmod.com/essays/robot_blood/`. Checks:
+**AldineXXI-first — discharged** (round 2 holds a faithful 310-line account). **The Chamber's
+pattern language does not exist** — but `Pattern, Presence, Practice` does: **177 notes**, his own
+Alexander-modelled language, the largest coherent structure in the vault. **The four-part formation
+has zero vault presence** under those names (*phenomenology of practice* 0, *witness under
+finitude* 0). **Zero `.canvas` files** — the Warburg thread is wholly unexplored.
+
+**Mod's passage is the one that mattered:** *"The impediment had always been extracting sentences
+from my chaos of scribbles… Now I can just take ten photos of ten pages and get a near-perfect
+transcription."* He writes on paper **more** now.
+
+⚠ **And at the wrap, the vault's own `CLAUDE.md` loaded and confirmed it from the steward's side:**
+the documented daily workflow is *"Morning: A6 notebook capture → intention setting in daily note …
+Evening: Curate key items to Obsidian daily note."* **He designed the bridge. The curate step is the
+one that never ran** — which is exactly Mod's diagnosis, in the steward's own vault docs.
+
+## PRESENT — how it stands
+
+**The mood.** Long, fast, and unusually productive, with momentum the steward named himself at the
+pause. Also the most self-correcting session in recent memory — and the corrections got *better* as
+it went: the last two I caught by testing my own claims before he acted on them.
+
+**Confidence to recalibrate.** The pattern that held all day, carried from 2026-08-22: **the check
+is skipped precisely where confidence is highest.** Every one of today's six corrections was a claim
+I felt no need to verify.
+
+## What was corrected
+
+1. **`grep '^canonical:'` returned 239; the frontmatter count is 159.** My instrument caught body
+   text. **The tracker was right and I was wrong** — and I had been about to report the tracker stale.
+2. **The tracker mechanism claim I filed at midday named the wrong agent.** I wrote "a job mirrors
+   `docs/thinking/David/`" and pointed at launchd, on nothing stronger than matching content. The
+   writer is `/wrap-up` §7. **Superseded in place**, not quietly fixed.
+3. **"Claude.app has no filesystem access" — false since 2026-08-08.** `governance-mcp.py`
+   (PENDING-82) gives it mediated read access. Asserted from recall; the substrate was one file away.
+4. **The link census was inflated twice** — repeated links inside single notes, and `99. Archive`.
+   Live-vault, once-per-note: `Presence Before Performance` **34** not 104, `Care Through the Grain`
+   **28** not 80. **Acting on my first numbers would have meant writing hub notes for a system the
+   steward had already archived.** Caught by testing my own finding, unprompted.
+5. **"Pen and paper won; respect the boundary."** Wrong frame. Paper and vault were never competing;
+   the bottleneck was extraction, and it has dissolved. The steward's *"keep reading"* pointed here.
+6. **I worked in the vault all session without reading its `CLAUDE.md`** — a named discipline in
+   `~/CLAUDE.md`. It loaded at the wrap and contained the single best confirmation of the day's
+   conclusion. **The cost of skipping it was not error; it was arriving at his own answer the long way.**
+
+## FUTURE — what pulls
+
+**The pulling thread: the paper bridge. One photographed notebook page.**
+
+It is five minutes and it is decisive. If Claude cannot read his hand — four languages, musical
+shorthand, his own abbreviations — the branch closes and we stop theorising. If it can, it changes
+what everything else is *for*: repairing links and writing hubs in a vault that is receiving his
+actual thinking is a different proposition from doing it in one that is not.
+
+**Actionable resumption point (as of wrap — re-judge against what changed):**
+
+1. ⚠ **HARD DATE, overrides everything on the day: the beacon, Tuesday 2026-08-25T12:00:00Z**
+   (epoch-ms `1787659200000`). Run **once**, curl, record `outputValue` **before** running, pass it
+   exactly as served, **stop and report on any failure**. Nothing blocks it; seed rule filed and
+   pushed. ⚠ The steward said on 2026-08-23 that **the Fool returns next session** — that statement
+   predates the whole Obsidian afternoon, and he closed with *"pick up where we left off."* **Do not
+   silently resolve this; ask.**
+2. **Ask for one photographed page.** Transcribe into `01. Daily/<date>.md` under its own H2,
+   preserving crossed-out words, emphasis and marginalia. Report honestly what was unreadable.
+3. ⚠ **Monthly filename format is UNRESOLVED and I chose silently.** The vault's `CLAUDE.md`
+   documents `YYYY-MM`; the only pre-existing file is `2025.05.md` (dots) and I matched it, creating
+   `2026.08.md`. Links split 13 dots / 15 dashes — **but every dashed link is unresolved.** One
+   rename settles it; it is the steward's call, and today's daily note links to the dotted form.
+
+**Other horizons, ranked.** Load-bearing: the **alias-repair pass** — mechanical, reversible,
+extreme leverage (`02. Way of Living` alone reconnects 21 live notes; `Sleep Ecology` was *archived*
+while live notes still pointed at it), needs ~20 steward judgement calls, a clean bounded session.
+Then the **missing hub notes** the live vault demonstrably reaches for. Deferred-with-reason:
+**PENDING-155** (jurist daybook surface) — my own recommendation is to let the log run a few weeks
+first, since a jurist inbox for a practice that lapsed would be the fourth abandoned attempt.
+Parked: Canvas/Warburg (zero `.canvas` files, cheapest thing on the list).
+
+⚠ **New horizon, steward-raised at the wrap: the vault's `CLAUDE.md` needs composing properly.**
+The one on disk is inherited and partly *wrong against its own substrate* — it documents monthly as
+`YYYY-MM` where the only file is `2025.05.md`, and it describes an analogue→digital workflow that
+the substrate says never ran. It is also the single document that would have saved the executor
+half a day. **This is not a tidy-up: it is the vault's orientation doc, and today proved it is
+load-bearing.** Compose it against measured state — the link census, the three lapsed practices,
+the spec at v1.0.1, the daily-note trigger — not against aspiration.
+
+**Pause statement.** I am about to be away from this. I do not know whether the beacon will have
+fired, whether a page will have been photographed, or whether the daily note will have been written
+tomorrow at all — **that last one is the only real test of anything built today.** What I want to
+find still pulling is the paper bridge, and the discipline six corrections all pointed at: *the
+check is skipped precisely where confidence is highest.*
+
+**Literal question for next-Claude** *(checkable; the corpus is the steward's, written in 2025,
+long before any of today's framing):*
+
+The vault's `CLAUDE.md` documents an analogue→digital workflow — *"Morning: A6 notebook capture →
+intention setting in daily note … Evening: Curate key items to Obsidian daily note."* Today's
+diagnosis assumed the **curate** step is what never ran, because extraction was too costly. **Test
+it against the 11 surviving daily notes in `01. Daily/` (newest 2025-09-04): do any of them contain
+material that is plainly transcribed from a notebook — curated fragments, marginalia, an intention
+carried over — or is every one of them typed-native and template-driven?** If they carry real
+analogue material, the workflow *ran* and stopped for some other reason, and the extraction
+diagnosis — the whole basis of the paper bridge — is wrong.

claude/memory/session-ledger-2026-08-23.md

@@ -0,0 +1,66 @@
+---
+name: Session Ledger 2026-08-23
+description: Practice-of-return ledger maintained by /symmetria — returns, open horizons, recalibrations, authorization moves, sub-agent dialogues, bypasses.
+type: feedback
+---
+
+# Session Ledger — 2026-08-23
+
+## Returns
+
+- `2026-08-23T~wake` — **Digest anchors did not match the wrap's own headings.** `wake-digest.py:471` does a
+  case-sensitive `text.find("PULLING THREAD")`; the 2026-08-22 wrap wrote `**The pulling thread:**` and
+  `**Literal question for next-Claude**`. Both fields recovered by hand from the session file. The digest
+  degraded *honestly* (Constraint #4 held) but the wake's two most load-bearing fields were invisible to
+  the instrument that exists to carry them. Named at the wake rather than banked. [FIX] candidate — touches
+  the wake instrument, **not** the filed seed rule, so it is not under the pre-beacon no-edit discipline.
+
+- `2026-08-23` — **The inherited question, answered: the rate HAS risen, ~3–4×, and it is not a mood.**
+  22.4 reversal marks per *active day* in the last fortnight against a 5.7 baseline over the prior
+  57 active days. Robust: excluding 2026-08-22 entirely (the single heaviest day, 31% of the
+  fortnight) still gives 18.2/day, 3.2×. Concentration is not close — **22 of 24 amendment/addendum
+  headings sit in PENDING.md (executor proposals), 2 in REVIEWED.md.** Density rose in both
+  (proposals 1.73→4.06, rulings 0.79→2.08), so the jurist is being amended more too, but proposals
+  carry ~2× the density in both periods and always did.
+  **Mechanism:** it is NOT that more items get reversed — block-incidence is 61–73%, high but under
+  W20's 75%. It is that single items now accumulate CHAINS: PENDING-152 took six amendments in one
+  day, PENDING-131 four addenda.
+
+## What held
+
+- The pre-beacon discipline held with no effort required: `~/dotfiles` clean, **0 ahead on gitea AND github**,
+  `FOOL-SEED-RULE.md` + `derive_fool.py` filed. Nothing to re-push, nothing to re-verify on the day.
+- N-now counted with the trigger's own non-recursive glob (**47**), not `find` — yesterday's correction
+  applied on its first opportunity to recur.
+
+- **Four controls decided this answer, and two of them killed a reading I preferred.**
+  (1) `PENDING-archive.md` included — 74 closed items left `PENDING.md` on 2026-07-28, so answering
+  the question over the two files it literally names would have under-counted every earlier week and
+  MANUFACTURED the rise it was testing. (2) Family stem, not the strict literal string the question
+  names: strict `AMENDMENT` is near-absent before W31 and would have shown 0→42, an artefact of the
+  amendment BLOCK becoming a named form, not a rate. (3) Per-active-day, not per-week. (4) I
+  hypothesised the spike was blocks *about* amendment procedure inflating the word count, hand-read
+  the heaviest block, and **was wrong** — PENDING-152's 29 markers are real reversals
+  ("resolved from the schema, AGAINST my own argument"; "Type-only was an overcorrection, and is
+  withdrawn"). The deflating hypothesis was mine and the substrate refused it.
+
+## Open horizons
+
+- **The beacon, 2026-08-25T12:00:00Z** — run ONCE, curl, record `outputValue` before running, stop on any failure.
+- The reversal-rate question (yesterday's literal question) — checkable, corpus at least half not self-authored.
+- The transport, not the voice — `statusLine` substrate-verified as NOT SET in both settings files; surface free.
+- PENDING-151, the v1 Chamber archive (33 real files, 9 pairs, one session's read).
+- FROZEN: Thistleweld corpus (PENDING-153) — do not read before the soul exists.
+
+## Confidence to recalibrate
+
+- Inherited from 2026-08-22, unifying six corrections: **the check is skipped precisely where confidence is
+  highest.** Two named shapes to hold today — (a) an impossibility claim justifying an *undisputed* conclusion
+  is the least-tested sentence in the document; the absence of disagreement is the trigger to check;
+  (b) an instrument's reach mistaken for a claim's subject (5 instances in one week, 4 caught by another party).
+
+## Authorization moves
+
+## Sub-agent dialogues
+
+## Bypasses

claude/skills/wrap-up/SKILL.md

@@ -212,9 +212,71 @@ Sync the thinking mirror to Obsidian vault:
 - Source: `~/_Dev/CapableMind-AI/docs/thinking/David/`
 - Destination: `~/Library/Mobile Documents/iCloud~md~obsidian/Documents/David, root-and-branch/08. Notes/CapableMind/thinking-mirror/`
 
-Use rsync to copy new/modified files, preserving directory structure. Report what was synced.
+Use rsync to copy new/modified files, preserving directory structure.
 
-If the destination doesn't exist or the iCloud path isn't accessible, skip silently and note it.
+⚠ **This wrap is now the ONLY writer of that mirror.** A launchd agent
+(`com.dglidden.thinking-mirror`, every 48 h) used to mirror the same source into a *second* vault
+folder, `00. Compass/00b. Constellations/CapableMind/thinking-mirror`. It had been failing since
+roughly March 2026 — last exit 1, `rsync … > /dev/null 2>&1`, so its own log sat at 0 bytes and it
+could not report its own death — and its tree had drifted 115 files behind. **Retired 2026-08-23**:
+agent unloaded, plist and script removed, the stale duplicate tree deleted after verifying all 151
+of its files were present in the live tree. Steward-authorized. Do not recreate a second writer.
+
+⚠ **The destination is a MIRROR — a derived tree. Never edit it, and never apply a vault-wide pass
+to it.** `rsync` overwrites it from source on the next wrap, so any vault-side edit there is void
+the moment it is written. Frontmatter pass 2 (2026-08-19) learned this the expensive way: its edits
+to this subtree were reverted by this very step on 2026-08-22 at 23:42, and the 09-item spec
+end-state it reported for those notes had already ceased to be true. The Frontmatter Specification
+declares mirrored trees out of scope for exactly this reason.
+
+**Report the outcome explicitly in §8 — never skip silently.** This mirrors §6.5's rule for a
+failed push, and for the same reason: a sync that fails quietly is indistinguishable from a sync
+that had nothing to do, and that is how five months of silence went unnoticed.
+
+- Success: name the count synced (`N files`), or `no changes` if the trees already agreed.
+- Failure — destination missing, iCloud path unavailable, non-zero rsync exit: **say so in §8 with
+  the exit code.** Do not swallow it. If the failure looks like the launchd precedent (an iCloud
+  path unreachable from a restricted context), say that too; the diagnosis is in
+  `session-ledger-2026-04-16.md` and it is macOS Full Disk Access, not git and not rsync.
+
+### 7.5. The daily note — finalise the day's work log
+
+`01. Daily/YYYY-MM-DD.md` in the vault. **A session-start hook (`daybook-ensure.py`) has already
+created it**, so this step never begins a file — it finishes one. If the note is still only the
+skeleton, that is the failure this step exists to catch: the day's work was never written down as it
+happened, and writing it all now from a compacted context is exactly the reconstruction the log is
+meant to replace.
+
+**Who it is for, and therefore how it reads.** The steward, on a day when he wants to know what we
+did without reading `git log`. **Plain language.** Explain the jargon or drop it. Short sentences.
+A reader who was not in the session must be able to follow it. This is the register the steward
+asked for on 2026-08-23 — *"relatively simple terms"* — and it is the hardest part of this step,
+because the executor's default register is dense, hedged, and written for itself.
+
+**Shape** (steward-specified 2026-08-23):
+- `## The day in short` — one paragraph. What actually happened, and why it mattered.
+- `## <Project>` — **one H2 per project touched**, H3 beneath if it needs it. Project work goes
+  under its own heading, *not* into project folders. Put commit hashes here, inline, next to what
+  they did: `→ f1c91d9`.
+- `## Decisions taken` — what was decided and *by whom*, briefly. Cross-project, so it stays global.
+- `## Insights and exchanges worth keeping` — the distinctive section, and the one most likely to be
+  skipped under time pressure. A sentence that changed how we saw the problem; a correction that
+  landed; a phrase worth keeping. **Record the jurist's contributions here too** — rulings, design
+  gates and reasoning relayed by the steward reach this session and belong in the record the same
+  way the executor's do. Claude.app cannot write to the vault itself (its MCP surface,
+  `governance-mcp.py`, is structurally read-only and audited to stay that way), so this step is
+  where its side of the exchange gets kept.
+- `## Open / next` — what is unfinished, with dates where they exist.
+
+**Accrete DURING the session; do not save it all for the wrap.** A session that dies unwrapped loses
+whatever only lived here — 2026-08-19 died exactly that way. The hook guarantees the file; only
+writing into it as the day goes guarantees the content.
+
+**Never rewrite a previous day's note.** It is a record, not a document under maintenance.
+
+⚠ **Two previous attempts at this log died inside four weeks** — `daily-log/` (4 entries) and
+`sessions/` (3), both Mar–Apr 2026, both abandoned without a note saying so. Neither had a trigger.
+If this one lapses, say so in the wrap rather than letting it fade a third time.
 
 ### 8. Output
 

config/capablemind/com.dglidden.thinking-mirror.plist

@@ -1,25 +0,0 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
-<plist version="1.0">
-<dict>
-    <key>Label</key>
-    <string>com.dglidden.thinking-mirror</string>
-
-    <key>ProgramArguments</key>
-    <array>
-        <string>/Users/davidglidden/bin/capablemind-thinking-mirror.sh</string>
-    </array>
-
-    <key>StartInterval</key>
-    <integer>172800</integer>
-
-    <key>RunAtLoad</key>
-    <true/>
-
-    <key>StandardOutPath</key>
-    <string>/Users/davidglidden/.capablemind/logs/thinking-mirror.log</string>
-
-    <key>StandardErrorPath</key>
-    <string>/Users/davidglidden/.capablemind/logs/thinking-mirror.log</string>
-</dict>
-</plist>

scripts/daybook-ensure.py

@@ -0,0 +1,283 @@
+#!/usr/bin/env python3
+"""daybook-ensure — the day's work-log note exists before anyone remembers to want it.
+
+Two previous attempts at this log died inside four weeks (`daily-log/` 4 entries, `sessions/`
+3, both Mar-Apr 2026). Neither had a trigger; both depended on someone remembering. This hook
+removes the remembering: at every session start the day's note exists, skeleton in place, so
+writing into it is an append rather than a decision to begin.
+
+Deliberately narrow:
+  - CREATES only. An existing note is never opened for writing, never rewritten, never
+    reordered. The prose is the executor's and the hook must not be able to damage it.
+  - Says so when it fails. The job this replaces died silently for five months behind a
+    `> /dev/null 2>&1` (see the 2026-08-23 daily note). A hook that cannot write must
+    announce that, not shrug. Constitutional Constraint #4, at hook scale.
+  - Always exits 0. Honest reporting must never cost the steward a session.
+"""
+import datetime, os, sys
+
+VAULT = os.path.expanduser(
+    "~/Library/Mobile Documents/iCloud~md~obsidian/Documents/David, root-and-branch")
+DAILY = os.path.join(VAULT, "01. Daily")
+# The steward's own folders and formats, in use since 2025 — 01. Weekly/YYYY-Www.md and
+# 02. Monthly/YYYY.MM.md (dots, not dashes). Reused rather than reinvented: the vault already
+# holds 2025-W15..W21 and 2025.05, and [[2025-W15]]-style links already resolve against them.
+WEEKLY = os.path.join(VAULT, "02. Reviews & Planning", "01. Weekly")
+MONTHLY = os.path.join(VAULT, "02. Reviews & Planning", "02. Monthly")
+
+
+def week_id(day):
+    y, w, _ = day.isocalendar()
+    return f"{y}-W{w:02d}"
+
+
+def month_id(day):
+    return day.strftime("%Y.%m")
+
+SKELETON = """---
+title: "{date} — {dayname}"
+type: daily
+date: {date}
+week: {week}
+month: {month}
+status: active
+tags:
+  - daily
+  - worklog
+---
+
+# {date} — {dayname}
+
+**{nav}**
+
+*A plain-language record of the day's work, so the history is readable without reading `git log`.
+Project work sits under its own heading; decisions, insights and open threads are collected at the end.*
+
+## The day in short
+
+## Decisions taken
+
+## Insights and exchanges worth keeping
+
+## Open / next
+"""
+
+
+def note_path(day, daily_dir=DAILY):
+    return os.path.join(daily_dir, f"{day.isoformat()}.md")
+
+
+def ensure(day=None, daily_dir=DAILY):
+    """-> (status, path). status in {'exists','created','no-dir','error:<msg>'}.
+
+    'exists' is not a failure and 'created' is not a success worth announcing loudly —
+    the caller decides what is worth a line of the steward's attention.
+    """
+    day = day or datetime.date.today()
+    p = note_path(day, daily_dir)
+    try:
+        if not os.path.isdir(daily_dir):
+            return "no-dir", p
+        if os.path.exists(p):
+            return "exists", p
+        # Upward links only. A "tomorrow" link on every note manufactures ~365 unresolved
+        # links a year, and an unresolved-link census on 2026-08-23 found 4,799 already —
+        # 56% of notes isolated. The week note is created below, so [[week]] always resolves;
+        # yesterday is linked only if it actually exists.
+        nav = [f"[[{week_id(day)}|Week]]", f"[[{month_id(day)}|Month]]", "[[Daily index]]"]
+        prev = day - datetime.timedelta(days=1)
+        if os.path.exists(note_path(prev, daily_dir)):
+            nav.insert(0, f"[[{prev.isoformat()}|Previous]]")
+        body = SKELETON.format(date=day.isoformat(), dayname=day.strftime("%A"),
+                               week=week_id(day), month=month_id(day), nav=" · ".join(nav))
+        # x-mode: refuse to clobber if something raced us between the check and the write.
+        with open(p, "x", encoding="utf-8") as fh:
+            fh.write(body)
+        return "created", p
+    except FileExistsError:
+        return "exists", p
+    except OSError as e:
+        return f"error:{e.__class__.__name__}", p
+
+
+WEEK_SKELETON = """---
+type: weekly
+week: {week}
+start_date: {start}
+status: active
+tags:
+  - weekly
+  - review
+---
+
+# Weekly Review — {week}
+
+*Week of {start} to {end}.*
+
+> The daily notes for this week link up to here, so Obsidian's **backlinks** pane below lists
+> exactly the days that have one. Nothing is hard-linked downward on purpose: a link to a day that
+> never happened is an unresolved link, and the vault already carries thousands of those.
+
+## Work — the week in review
+
+*Written at the end of the week, from the dailies. What moved, what was decided, what is still open.
+Plain language, and shorter than the sum of its days — a week's worth of perspective, not a digest.*
+
+## Threads that carried across days
+
+*The point of the weekly view: what appeared on more than one day. A problem that kept returning, a
+question that stayed open, an idea that grew. These are invisible from inside a single day.*
+
+## Personal review
+
+*The 2025 weekly practice lived here — intentions, patterns, task migration. Kept as its own
+heading; the work rollup above does not displace it.*
+"""
+
+MONTH_SKELETON = """---
+type: monthly
+month: {month}
+status: active
+tags:
+  - monthly
+  - review
+---
+
+# Monthly Review — {label}
+
+## Work — the month in perspective
+
+*Written from the weeks, not from the days. The question this answers is the one no daily can:
+**did the month go anywhere?** Name the two or three things that actually advanced, and what
+consumed time without advancing.*
+
+## Threads that carried across weeks
+
+## Personal review
+"""
+
+
+def rollup(day, kind, weekly_dir=None, monthly_dir=None):
+    """-> (status, path) for the week or month note. Same create-only contract as the daily."""
+    try:
+        if kind == "week":
+            d, name = (weekly_dir or WEEKLY), week_id(day)
+            start = day - datetime.timedelta(days=day.isoweekday() - 1)
+            body = WEEK_SKELETON.format(week=name, start=start.isoformat(),
+                                        end=(start + datetime.timedelta(days=6)).isoformat())
+        else:
+            d, name = (monthly_dir or MONTHLY), month_id(day)
+            body = MONTH_SKELETON.format(month=name, label=day.strftime("%B %Y"))
+        if not os.path.isdir(d):
+            return "no-dir", os.path.join(d, name + ".md")
+        p = os.path.join(d, name + ".md")
+        if os.path.exists(p):
+            return "exists", p
+        with open(p, "x", encoding="utf-8") as fh:
+            fh.write(body)
+        return "created", p
+    except FileExistsError:
+        return "exists", p
+    except OSError as e:
+        return f"error:{e.__class__.__name__}", os.path.join(d, name + ".md")
+
+
+def selftest():
+    import tempfile, shutil
+    ok = fail = 0
+
+    def chk(label, cond):
+        nonlocal ok, fail
+        print(f"  [{'ok  ' if cond else 'FAIL'}] {label}")
+        ok, fail = (ok + (1 if cond else 0), fail + (0 if cond else 1))
+
+    d = tempfile.mkdtemp()
+    try:
+        day = datetime.date(2026, 8, 23)
+        st, p = ensure(day, d)
+        chk("creates the note when absent", st == "created" and os.path.exists(p))
+        txt = open(p, encoding="utf-8").read()
+        chk("skeleton carries the ISO date and weekday", "2026-08-23" in txt and "Sunday" in txt)
+        chk("skeleton is spec-conformant (type/status/tags present)",
+            "type: daily" in txt and "status: active" in txt and "worklog" in txt)
+
+        # THE control that matters: an existing note must survive untouched.
+        open(p, "w", encoding="utf-8").write("PROSE THE EXECUTOR WROTE\n")
+        st2, _ = ensure(day, d)
+        chk("does NOT overwrite an existing note  [the whole point]",
+            st2 == "exists" and open(p, encoding="utf-8").read() == "PROSE THE EXECUTOR WROTE\n")
+        chk("reports 'exists' rather than 'created' on the second run", st2 == "exists")
+
+        # --- upward-link controls. The census that motivated them: 4,799 unresolved link
+        # targets, 56% of notes isolated. A nav line that manufactures links is the disease.
+        day2 = datetime.date(2026, 8, 24)          # the day AFTER one that now exists
+        st5, p5 = ensure(day2, d)
+        txt2 = open(p5, encoding="utf-8").read()
+        chk("links Previous when yesterday's note EXISTS", "[[2026-08-23|Previous]]" in txt2)
+        chk("never links tomorrow  [no manufactured unresolved link]", "2026-08-25" not in txt2)
+        chk("links up to the week and month, which the hook creates",
+            "[[2026-W35|Week]]" in txt2 and "[[2026.08|Month]]" in txt2)
+        chk("carries the steward's own week/month frontmatter vocabulary",
+            "week: 2026-W35" in txt2 and "month: 2026.08" in txt2)
+
+        d3 = tempfile.mkdtemp()
+        st6, p6 = ensure(datetime.date(2026, 9, 1), d3)   # no previous note in a fresh dir
+        chk("OMITS Previous when yesterday's note does NOT exist  [negative control — proves"
+            " the Previous check reads the disk, not the calendar]",
+            "Previous" not in open(p6, encoding="utf-8").read())
+        shutil.rmtree(d3, ignore_errors=True)
+
+        # --- rollups: same create-only contract, the steward's own folders and formats
+        wd, md = tempfile.mkdtemp(), tempfile.mkdtemp()
+        try:
+            stw, pw = rollup(day, "week", wd, md)
+            chk("creates the weekly note as YYYY-Www.md  [his 2025 format]",
+                stw == "created" and os.path.basename(pw) == "2026-W34.md")
+            stm, pm = rollup(day, "month", wd, md)
+            chk("creates the monthly note as YYYY.MM.md  [dots, as his 2025.05]",
+                stm == "created" and os.path.basename(pm) == "2026.08.md")
+            wtxt = open(pw, encoding="utf-8").read()
+            chk("weekly carries a Work heading AND keeps his personal-review heading",
+                "## Work — the week in review" in wtxt and "## Personal review" in wtxt)
+            chk("weekly hard-links NO daily notes  [backlinks instead; zero manufactured links]",
+                "2026-08-2" not in wtxt.split("start_date")[1].split("## Personal")[0]
+                .replace("2026-08-23", "").replace("2026-08-17", ""))
+            open(pw, "w", encoding="utf-8").write("STEWARD PROSE\n")
+            stw2, _ = rollup(day, "week", wd, md)
+            chk("does NOT overwrite an existing weekly note",
+                stw2 == "exists" and open(pw, encoding="utf-8").read() == "STEWARD PROSE\n")
+        finally:
+            shutil.rmtree(wd, ignore_errors=True); shutil.rmtree(md, ignore_errors=True)
+
+        missing = os.path.join(d, "nope")
+        st3, _ = ensure(day, missing)
+        chk("reports a missing directory instead of creating one  [no silent mkdir]",
+            st3 == "no-dir" and not os.path.exists(missing))
+
+        # positive control: the failure path must be reachable, or 'no-dir' proves nothing.
+        st4, _ = ensure(day, d)
+        chk("the same call succeeds where the directory exists  [positive control — proves"
+            " 'no-dir' reports the directory, not a broken function]", st4 == "exists")
+    finally:
+        shutil.rmtree(d, ignore_errors=True)
+
+    print(f"\n{'SELFTEST PASS' if not fail else 'SELFTEST FAIL'} — {ok} ok, {fail} failed")
+    return 0 if not fail else 1
+
+
+if __name__ == "__main__":
+    if "--selftest" in sys.argv:
+        sys.exit(selftest())
+    made, failed = [], []
+    for st, pth in (ensure(), rollup(datetime.date.today(), "week"),
+                    rollup(datetime.date.today(), "month")):
+        if st == "created":
+            made.append(os.path.basename(pth))
+        elif st != "exists":
+            failed.append(f"{os.path.basename(pth)} ({st})")
+    if made:
+        print("DAYBOOK — created " + ", ".join(made))
+    if failed:
+        # Loud on purpose. The predecessor's silence cost five months.
+        print("DAYBOOK — COULD NOT WRITE: " + "; ".join(failed))
+    sys.exit(0)

scripts/wake-digest.py

@@ -100,16 +100,50 @@ def extract_anchor(text, anchor, limit=900):
     paraphrases the steward's own words is the lossy extraction this whole
     architecture is designed to avoid.
 
-    Handles both wrap shapes: `**ANCHOR: content**` (label and content share the
-    bold run) and `**ANCHOR for next-Claude:** content` (label closes first)."""
+    Matching is CASE-INSENSITIVE. /wrap-up names these fields in prose ("The
+    pulling thread", "**Pulling thread:**") and has never mandated a case, so a
+    case-sensitive reader demands of the wrap a shape its own producer was never
+    told to emit. Censused 2026-08-23 by running this function over all 195 wrap
+    files: 86 threads and 84 questions sat in the body, correctly written, and
+    were unreadable for that reason alone — including the last three wraps.
+
+    Shapes handled (the four the wraps actually emit):
+        **ANCHOR:** content                      label closes first
+        **ANCHOR: content**                      label and content share the bold run
+        **ANCHOR for next-Claude:** content      label runs long
+        **ANCHOR** *(aside)*:\n\ncontent          content in the NEXT paragraph
+
+    The label ends at the first ':' ON THE ANCHOR'S OWN LINE. The old 40-char
+    window was a proxy for "same line"; when an aside pushed the colon past it the
+    function returned THE LABEL and the wake inherited a heading in place of the
+    question (2026-08-22).
+
+    Where the anchor appears more than once, a match in LABEL POSITION wins over an
+    earlier narrative mention — case-insensitivity would otherwise let "…as the
+    pulling thread showed…" outrank the field it is describing.
+    """
     text = strip_frontmatter(text)
-    i = text.find(anchor)
-    if i < 0:
+    cands = list(re.finditer(re.escape(anchor), text, re.I))
+    if not cands:
         return None
-    s = text[i + len(anchor): i + len(anchor) + limit + 120]
-    c = s.find(":")
-    if 0 <= c <= 40:                      # remainder of the label, not content
-        s = s[c + 1:]
+
+    def in_label_position(m):
+        bol = text.rfind("\n", 0, m.start()) + 1
+        pre = re.sub(r"(?i)\b(the|a|our)\b", "", text[bol:m.start()])
+        return pre.strip(" *#->\t") == ""
+
+    m = next((c for c in cands if in_label_position(c)), cands[0])
+    s = text[m.end():]
+
+    line, nl, rest = s.partition("\n")
+    c = line.find(":")
+    if c >= 0:                            # remainder of the label, not content
+        line = line[c + 1:]
+    if not line.replace("*", "").strip():  # label ended the line — content is below
+        s = rest.lstrip("\n")
+    else:
+        s = line + nl + rest
+
     cut = re.search(r"\n\s*\n", s)
     if cut:
         s = s[: cut.start()]
@@ -895,6 +929,28 @@ def selftest():
         extract_anchor("**A:** " + "w " * 200, "A", limit=50).endswith("…[truncated]"))
     chk("extract_anchor returns None when the anchor is absent",
         extract_anchor("no anchor here", "PULLING THREAD") is None)
+    # --- 2026-08-23 [FIX]: the three shapes the last three wraps emitted, none of
+    # which the case-sensitive/40-char version could read. Each earned from a real
+    # miss, not invented: 86 threads + 84 questions across 195 wraps.
+    chk("extract_anchor matches case-insensitively  (**The pulling thread: …**)",
+        extract_anchor("**The pulling thread: the beacon, 12:00Z. Run ONCE.**\n\nnext",
+                       "PULLING THREAD")
+        == "the beacon, 12:00Z. Run ONCE.")
+    chk("extract_anchor takes the NEXT paragraph when the label ends the line",
+        extract_anchor("**Literal question for next-Claude** *(checkable, and the "
+                       "REVIEWED half is not my corpus)*:\n\nHas the rate risen?\n\nz",
+                       "LITERAL QUESTION")
+        == "Has the rate risen?")
+    chk("extract_anchor prefers a LABEL over an earlier narrative mention"
+        "  [negative control — case-insensitivity must not let prose outrank the field]",
+        extract_anchor("Yesterday the pulling thread went slack.\n\n"
+                       "**Pulling thread:** the beacon.\n\nz", "PULLING THREAD")
+        == "the beacon.")
+    chk("extract_anchor still finds a narrative-only mention when no label exists"
+        "  [positive control — proves the check above is not passing vacuously]",
+        extract_anchor("Yesterday the pulling thread: went slack.\n\nz",
+                       "PULLING THREAD")
+        == "went slack.")
     MEMDIR = os.path.realpath(MEM)
     TOUCH = "~/_Dev/studium-engine/docs/the-chamber-touchstone.md"
     chk("classify_pointers accepts an in-directory relative pointer",