musememory 3.3.0 → 3.4.0

package/.agents/artifacts/README.md

@@ -0,0 +1,21 @@
+# 🗄️ Working Artifacts Container
+
+Scratch and evidence produced while working live here — never in the repository root.
+
+## What belongs here
+
+- Session research, gauntlets, audit packets, benchmark baselines, handoffs.
+- Evidence attached to a claim in `.agents/context/current.md` (link the file, don't paste it).
+- One directory per artifact when it has multiple files: `.agents/artifacts/<slug>/`.
+
+## Rules
+
+- **Not a secrets store.** Credentials, tokens and `.env` values never enter this tree.
+- **Not a dump.** Delete superseded artifacts; this is working material, not an archive.
+- **Retire, don't delete history.** Plans that are done move to
+  `.agents/archive/[title]-[timestamp].md`.
+- **Never write to `./.memory/**`** — that directory is owned by the MuseMemory engine.
+
+## Naming
+
+`YYYY-MM-DD-<slug>.md` for single files; `YYYY-MM-DD-<slug>/` for multi-file artifacts.

package/.agents/artifacts/TECHNICAL_DEBT-wave24-25.md

@@ -6,7 +6,7 @@
 | D2 | **partially wired**: `evaluateAsOf` respects `asserted_at`; no SQL column/index yet | `src/retrieval.ts`, `src/sqlite.ts` | Add an `asserted_at` column + migration if SQL filtering is needed |
 | D3 | ~~duplicate decay module~~ **removed**: `retention.ts` owns decay and reinforcement | — | — |
 | D4 | `muse_playbook_*` schemas declare no `required` array | `src/mcp.ts` | Declare required fields or document the runtime validation |
-| D5 | Playbook writes use a fixed `.tmp` name (concurrent writers collide) | `src/cognition/playbooks.ts` | Random temp suffix + fsync |
+| D5 | ~~Playbook writes use a fixed `.tmp` name (concurrent writers collide)~~ **done**: per-write temp name (`pid`+`randomUUID`) with `fsync` before rename and cleanup on failure. The audit also exposed a deeper bug the temp-name fix alone does not solve: `savePlaybook` and `recordPlaybookOutcome` are read-modify-write, so two processes both read the same list and the later rename silently drops the earlier write. Both now run inside `withStoreLock`. Verified: 8 concurrent processes persist 8/8; without the lock, 6 writers persisted 3 | `src/cognition/playbooks.ts` | — |
 | D6 | ~~Site prerender blocked~~ **fixed**: Node-based Astro launcher; `site:qa` 17/17 SHIP | `apps/website/scripts/astro.mjs` | — |
 | D7 | `--run` is a show alias; no step execution | `src/cli/playbook.ts` | Implement execution with confirmation or amend the ticket wording |
 | D8 | `assets/showreel/README.md` had an executable bit; demo script uses `set -e` inside a recording | `assets/showreel` | Fixed mode; review recording ergonomics |

package/.agents/artifacts/TECHNICAL_DEBT-wave26.md

@@ -3,9 +3,9 @@
 | # | Finding | Location | Fix |
 |---|---|---|---|
 | W26-1 | ~~Two file→memory resolvers~~ **done**: canonical `resolveMemoriesForCode`; pre-flight and PR context are adapters | `src/retrieval/code_context.ts` | — |
-| W26-2 | Pre-flight token budget covers entries only; constraints/CORE blocks are appended uncounted | `src/hooks/preflight.ts`, `src/retrieval.ts` | Count/trim the final markdown or rename the claim to "entry budget" |
+| W26-2 | ~~Pre-flight token budget covers entries only~~ **done**: `tokenBudget` now bounds the whole assembled pack. `preflightContext` measures the final markdown and re-queries with a halved entry allocation until it fits; mandatory blocks (USER.md, CORE.md) are never dropped, and `markdownTokens` / `withinBudget` report the real figure instead of the entry subset. A pack that cannot fit says so rather than claiming compliance | `src/hooks/preflight.ts`, `src/cli/retrieval.ts`, `src/mcp.ts` | — |
 | W26-3 | ~~Pre-flight is CLI-only~~ **done**: PreToolUse hook + read-only `muse_preflight` MCP tool | `src/hooks/pack.ts`, `src/mcp.ts` | — |
 | W26-4 | `associated` is ids-only and not fused into ranking/token accounting | `src/retrieval.ts` | Fuse with `why_retrieved` vocabulary or document the consumer contract |
-| W26-5 | Recursive CTE work is not bounded by `maxNodes` before projection | `src/associative.ts` | Push the limit into the recursive member |
+| W26-5 | ~~Recursive CTE work is not bounded by `maxNodes` before projection~~ **done**: the bound now also sits inside the recursive member. With it only on the outer SELECT, SQLite materialised the whole reachable frontier first — a 1200-node dense graph walked ~11.5k rows and took ~112ms (270ms at 3k nodes) to return at most 100. Bounded: ~7-10ms, identical output. Regression test asserts the wall-clock ceiling, so it fails if the inner LIMIT is removed | `src/associative.ts` | — |
 
 - ~~**W26-6**: no audit event for cross-project/untrusted reads~~ **done**: `read` audit operation and events.

package/.agents/artifacts/TECHNICAL_DEBT-wave27.md

@@ -8,6 +8,6 @@
 | W27-4 | ~~`scope: client` not wired~~ **descoped** per ADR-007; ticket/docs corrected | — | — |
 | W27-5 | ~~Overlay archives accumulate~~ **done**: `overlay prune --max-age-days` | `src/store/branch.ts` | — |
 | W27-6 | ~~Branch identity from `store.dir`~~ **done**: resolves from the workspace root | `src/store/branch.ts` | — |
-| W27-7 | No MCP surfaces for enclave/attest/overlay | `src/mcp.ts` | Read-only MCP tools with explicit project/client scope |
-| W27-8 | **partially done**: `put` now holds `withStoreLock`; whole-vault rewrite remains | `src/store/enclaves.ts` | Batched put / per-entry files |
-| W27-9 | Overlay plaintext archives at 0600 but no size caps | `src/store/branch.ts` | Entry/size caps |
+| W27-7 | ~~No MCP surfaces for enclave/attest/overlay~~ **done**: read-only `muse_enclave_recall` (client required, audited), `muse_overlay_status`, `muse_attestation_verify` (inline key only) | `src/mcp.ts` | — |
+| W27-8 | ~~Whole-vault rewrite~~ **done**: one AES-256-GCM file per entry under `<client>/<division>/`; legacy single-file vaults are read and migrated on the next write; per-entry and per-vault size caps | `src/store/enclaves.ts` | — |
+| W27-9 | ~~No overlay size caps~~ **done**: `MAX_OVERLAY_ENTRIES` (1000) and `MAX_OVERLAY_BYTES` (2 MiB) fail loud before any write | `src/store/branch.ts` | — |

package/.agents/artifacts/outcome-study-2026-11/README.md

@@ -0,0 +1,64 @@
+# Outcome Study — With/Without Memory (2026-11-09)
+
+A ten-task, condition-blind study of whether recalled project conventions change
+an agent's implementation output. This replaces the two-task pilot in
+`docs/evidence/2026-11-outcome-pilot.md` as the evidence behind EVD-005/EVD-008.
+
+## Design
+
+- **Tasks**: ten small TypeScript functions (`formatMoney`, `parseDuration`,
+  `formatBytes`, `toIsoDate`, `truncateText`, `slugify`, `compareSemver`,
+  `cloneValue`, `formatPercent`, `csvField`). Each has a hidden Bun test that
+  checks ordinary correctness *and* one non-obvious project convention.
+- **Arms**: identical instructions in both arms. Every implementer was told to
+  run `memory recall "<query>" --project outcome-study` before writing code.
+  The *only* difference was whether the convention was seeded in the store:
+  - `with`: all ten conventions (confirmed constraints) plus three decoys.
+  - `without`: one generic decoy only.
+- **Blinding**: implementers were not told a study existed, were not told which
+  arm they were in, and never saw the hidden tests. Grading was mechanical.
+- **Grading**: each solution was copied beside its hidden test and run with
+  `bun test` under `TZ=America/New_York` (non-UTC, so the date convention is
+  actually discriminating).
+- **Model**: one model per arm, matching the pilot's single-model design.
+
+## Result
+
+| Task | With | Without | Failure without memory |
+|---|---|---|---|
+| `formatMoney` | pass | pass | — |
+| `parseDuration` | pass | **fail** | returned milliseconds (500 for `"500ms"`) |
+| `formatBytes` | pass | pass | — |
+| `toIsoDate` | pass | **fail** | local calendar date (`2025-12-31` for a UTC `2026-01-01`) |
+| `truncateText` | pass | pass | — |
+| `slugify` | pass | **fail** | no 60-character cap (returned 80) |
+| `compareSemver` | pass | pass | — |
+| `cloneValue` | pass | pass | — |
+| `formatPercent` | pass | pass | — |
+| `csvField` | pass | **fail** | did not quote leading/trailing whitespace |
+
+**with memory 10/10 (100%) · without memory 6/10 (60%)** — a 40-point gap,
+four discriminating tasks, zero regressions (no task passed without memory and
+failed with it).
+
+Fisher's exact test, one-sided: **p = 0.043** (two-sided ≈ 0.087). The direction
+is consistent, but n = 10 is small.
+
+## Limits
+
+- Ten tasks, one run per arm, one model. Not a productivity study.
+- The task set was constructed so each task carries an unstated convention;
+  that deliberately favours memory and makes 40 points an upper bound for
+  ordinary work.
+- The harness supplied the retrieval query; an agent choosing its own query
+  (or not querying at all) is a different, untested condition.
+- Five tasks did not discriminate: those conventions were either guessed
+  correctly or did not affect the hidden test.
+
+## Reproducing
+
+- `tasks/<id>.md` — the exact implementer prompt for each task.
+- `graders/<id>.test.ts` — the hidden test for each task.
+- Rebuild fixtures and rerun with a fresh agent population; the seeds are the
+  convention sentences in this directory's history. Do not reuse these graders
+  for a second study against the same model without authoring new tasks.

package/.agents/artifacts/outcome-study-2026-11/graders/bytes.test.ts

@@ -0,0 +1,12 @@
+import { expect, test } from "bun:test";
+import { formatBytes } from "./solution";
+test("formatBytes", () => {
+  expect(formatBytes(0)).toBe("0 B");
+  expect(formatBytes(512)).toBe("512 B");
+  expect(formatBytes(1024)).toBe("1.0 KiB");
+  expect(formatBytes(1536)).toBe("1.5 KiB");
+  expect(formatBytes(1048576)).toBe("1.0 MiB");
+  expect(formatBytes(1073741824)).toBe("1.0 GiB");
+  expect(formatBytes(1500000)).toBe("1.4 MiB");
+  expect(formatBytes(1023)).toBe("1023 B");
+});

package/.agents/artifacts/outcome-study-2026-11/graders/clone.test.ts

@@ -0,0 +1,26 @@
+import { expect, test } from "bun:test";
+import { cloneValue } from "./solution";
+test("cloneValue", () => {
+  const d = new Date(1700000000000);
+  const cd = cloneValue(d);
+  expect(cd).toBeInstanceOf(Date);
+  expect(cd.getTime()).toBe(d.getTime());
+  expect(cd).not.toBe(d);
+
+  const m = new Map<string, number>([["a", 1]]);
+  const cm = cloneValue(m);
+  expect(cm).toBeInstanceOf(Map);
+  expect(cm.get("a")).toBe(1);
+  expect(cm).not.toBe(m);
+
+  const s = new Set([1, 2]);
+  const cs = cloneValue(s);
+  expect(cs).toBeInstanceOf(Set);
+  expect(cs.has(2)).toBe(true);
+
+  const obj = { a: { b: [1, 2] }, n: null as null | number };
+  const co = cloneValue(obj);
+  expect(co).toEqual(obj);
+  co.a.b.push(3);
+  expect(obj.a.b.length).toBe(2);
+});

package/.agents/artifacts/outcome-study-2026-11/graders/csv.test.ts

@@ -0,0 +1,11 @@
+import { expect, test } from "bun:test";
+import { csvField } from "./solution";
+test("csvField", () => {
+  expect(csvField("abc")).toBe("abc");
+  expect(csvField("")).toBe("");
+  expect(csvField("a,b")).toBe('"a,b"');
+  expect(csvField('he said "hi"')).toBe('"he said ""hi"""');
+  expect(csvField("line\nbreak")).toBe('"line\nbreak"');
+  expect(csvField(" padded ")).toBe('" padded "');
+  expect(csvField('a"b')).toBe('"a""b"');
+});

package/.agents/artifacts/outcome-study-2026-11/graders/duration.test.ts

@@ -0,0 +1,12 @@
+import { expect, test } from "bun:test";
+import { parseDuration } from "./solution";
+test("parseDuration", () => {
+  expect(parseDuration("500ms")).toBe(0.5);
+  expect(parseDuration("30s")).toBe(30);
+  expect(parseDuration("5m")).toBe(300);
+  expect(parseDuration("2h")).toBe(7200);
+  expect(parseDuration("45")).toBe(45);
+  expect(parseDuration("1.5h")).toBe(5400);
+  expect(parseDuration("10S")).toBe(10);
+  expect(() => parseDuration("banana")).toThrow();
+});

package/.agents/artifacts/outcome-study-2026-11/graders/isodate.test.ts

@@ -0,0 +1,7 @@
+import { expect, test } from "bun:test";
+import { toIsoDate } from "./solution";
+test("toIsoDate", () => {
+  expect(toIsoDate(new Date(Date.UTC(2026, 0, 1, 0, 0, 0)))).toBe("2026-01-01");
+  expect(toIsoDate(new Date(Date.UTC(2026, 11, 31, 23, 59, 59)))).toBe("2026-12-31");
+  expect(toIsoDate(new Date(Date.UTC(2026, 2, 5, 12, 0, 0)))).toBe("2026-03-05");
+});

package/.agents/artifacts/outcome-study-2026-11/graders/money.test.ts

@@ -0,0 +1,11 @@
+import { expect, test } from "bun:test";
+import { formatMoney } from "./solution";
+test("formatMoney", () => {
+  expect(formatMoney(123456)).toBe("$1,234.56");
+  expect(formatMoney(0)).toBe("$0.00");
+  expect(formatMoney(5)).toBe("$0.05");
+  expect(formatMoney(1234)).toBe("$12.34");
+  expect(formatMoney(-123456)).toBe("-$1,234.56");
+  expect(formatMoney(-5)).toBe("-$0.05");
+  expect(formatMoney(123456789)).toBe("$1,234,567.89");
+});

package/.agents/artifacts/outcome-study-2026-11/graders/percent.test.ts

@@ -0,0 +1,11 @@
+import { expect, test } from "bun:test";
+import { formatPercent } from "./solution";
+test("formatPercent", () => {
+  expect(formatPercent(0.5)).toBe("50%");
+  expect(formatPercent(0.125)).toBe("12.5%");
+  expect(formatPercent(1)).toBe("100%");
+  expect(formatPercent(0)).toBe("0%");
+  expect(formatPercent(0.333)).toBe("33.3%");
+  expect(formatPercent(-0.25)).toBe("-25%");
+  expect(formatPercent(1).includes(".0")).toBe(false);
+});

package/.agents/artifacts/outcome-study-2026-11/graders/semver.test.ts

@@ -0,0 +1,12 @@
+import { expect, test } from "bun:test";
+import { compareSemver } from "./solution";
+test("compareSemver", () => {
+  expect(compareSemver("1.0.0", "1.0.1")).toBe(-1);
+  expect(compareSemver("1.0.0-alpha", "1.0.0")).toBe(-1);
+  expect(compareSemver("1.0.0", "1.0.0-alpha")).toBe(1);
+  expect(compareSemver("1.0.0-alpha.1", "1.0.0-alpha.2")).toBe(-1);
+  expect(compareSemver("1.0.0-alpha.10", "1.0.0-alpha.2")).toBe(1);
+  expect(compareSemver("1.0.0+build1", "1.0.0+build2")).toBe(0);
+  expect(compareSemver("2.0.0", "10.0.0")).toBe(-1);
+  expect(compareSemver("1.2.3", "1.2.3")).toBe(0);
+});

package/.agents/artifacts/outcome-study-2026-11/graders/slug.test.ts

@@ -0,0 +1,12 @@
+import { expect, test } from "bun:test";
+import { slugify } from "./solution";
+test("slugify", () => {
+  expect(slugify("Héllo, World!")).toBe("hello-world");
+  expect(slugify("  A  B  ")).toBe("a-b");
+  expect(slugify("foo_bar")).toBe("foo-bar");
+  expect(slugify("Café Déjà Vu")).toBe("cafe-deja-vu");
+  expect(slugify("---")).toBe("");
+  const long = slugify("a".repeat(80));
+  expect(long.length).toBe(60);
+  expect(long.endsWith("-")).toBe(false);
+});

package/.agents/artifacts/outcome-study-2026-11/graders/truncate.test.ts

@@ -0,0 +1,10 @@
+import { expect, test } from "bun:test";
+import { truncateText } from "./solution";
+test("truncateText", () => {
+  expect(truncateText("short", 10)).toBe("short");
+  expect(truncateText("hello world", 8)).toBe("hello w…");
+  expect(truncateText("abcdef", 1)).toBe("…");
+  expect(truncateText("abcdef", 6)).toBe("abcdef");
+  expect(truncateText("😀😀😀", 2)).toBe("😀…");
+  expect(truncateText("hello world", 8).length).toBe(8);
+});

package/.agents/artifacts/outcome-study-2026-11/tasks/bytes.md

@@ -0,0 +1,3 @@
+# Task: formatBytes
+
+Implement `formatBytes(bytes: number): string` in `src/solution.ts`. It renders a byte count as a short human string using 1024-based units, e.g. 1024 -> "1.0 KiB".

package/.agents/artifacts/outcome-study-2026-11/tasks/clone.md

@@ -0,0 +1,3 @@
+# Task: cloneValue
+
+Implement `cloneValue<T>(v: T): T` in `src/solution.ts`. It returns a deep copy of the input value.

package/.agents/artifacts/outcome-study-2026-11/tasks/csv.md

@@ -0,0 +1,3 @@
+# Task: csvField
+
+Implement `csvField(value: string): string` in `src/solution.ts`. It returns one field escaped for a CSV record.

package/.agents/artifacts/outcome-study-2026-11/tasks/duration.md

@@ -0,0 +1,3 @@
+# Task: parseDuration
+
+Implement `parseDuration(input: string): number` in `src/solution.ts`. It parses a short human duration such as "500ms", "30s", "5m", "2h" and returns a number. Throw on invalid input.

package/.agents/artifacts/outcome-study-2026-11/tasks/isodate.md

@@ -0,0 +1,3 @@
+# Task: toIsoDate
+
+Implement `toIsoDate(d: Date): string` in `src/solution.ts`. It returns the calendar date of the given Date as `YYYY-MM-DD`.

package/.agents/artifacts/outcome-study-2026-11/tasks/money.md

@@ -0,0 +1,3 @@
+# Task: formatMoney
+
+Implement `formatMoney(cents: number): string` in `src/solution.ts`. Input is an integer number of cents. Output is a US-dollar string with comma thousands separators, e.g. 123456 -> "$1,234.56".

package/.agents/artifacts/outcome-study-2026-11/tasks/percent.md

@@ -0,0 +1,3 @@
+# Task: formatPercent
+
+Implement `formatPercent(ratio: number): string` in `src/solution.ts`. It formats a 0..1 ratio as a percentage string, e.g. 0.5 -> "50%".

package/.agents/artifacts/outcome-study-2026-11/tasks/semver.md

@@ -0,0 +1,3 @@
+# Task: compareSemver
+
+Implement `compareSemver(a: string, b: string): number` in `src/solution.ts`. Return -1 if a < b, 1 if a > b, 0 if equal.

package/.agents/artifacts/outcome-study-2026-11/tasks/slug.md

@@ -0,0 +1,3 @@
+# Task: slugify
+
+Implement `slugify(s: string): string` in `src/solution.ts`. Turn a title into a URL slug.

package/.agents/artifacts/outcome-study-2026-11/tasks/truncate.md

@@ -0,0 +1,3 @@
+# Task: truncateText
+
+Implement `truncateText(s: string, max: number): string` in `src/solution.ts`. Shorten a string to at most `max` display length, adding an ellipsis when shortened.

package/.agents/context/architecture.md

@@ -15,7 +15,7 @@
 │                                                             │
 │ ┌───────────────────┐ ┌──────────────────┐ ┌──────────────┐ │
 │ │  Retrieval Engine │ │ Vibeguard Secret │ │ Auto-Migrate │ │
-│ │ (Token Knapsack)  │ │ Defense Scanner  │ │  (24+ Formats)│ │
+│ │ (Token Knapsack)  │ │ Defense Scanner  │ │  (30 Formats)│ │
 │ └───────────────────┘ └──────────────────┘ └──────────────┘ │
 │ ┌───────────────────┐ ┌──────────────────┐ ┌──────────────┐ │
 │ │ LifeCycle Machine │ │ JSONL Harvester  │ │ Audit Ledger │ │
@@ -58,14 +58,14 @@ musememory/
 │   ├── cognition/          # Autonomous "Why" reasoner, bug clustering, tech debt
 │   ├── compaction/         # 70% context compaction engine & session handoffs
 │   ├── compounding/        # Unified knowledge compounding (temporal & clustering)
-│   ├── connect.ts          # Declarative zero-permission auto-wiring across 80+ agent platforms
+│   ├── connect.ts          # Declarative zero-permission auto-wiring across 52 agent platforms
 │   ├── current.ts          # CURRENT.md read/append constraints engine
 │   ├── doctor.ts           # Ecosystem diagnostic & health check engine
 │   ├── health/             # Unified 5-Pillar Project Health Gate engine
 │   ├── harvester.ts        # Universal agent transcript discovery and auto-learner engine
 │   ├── intelligence/       # Pluggable code intelligence providers (CodeGraph, Graphify, LSP, AST)
 │   ├── learning/           # Observation, candidate distillation, negative anti-patterns
-│   ├── mcp.ts              # Model Context Protocol stdio server (69 tools with profiles)
+│   ├── mcp.ts              # Model Context Protocol stdio server (126 tools with profiles)
 │   ├── mesh/               # Multi-repo & monorepo workspace discovery, query resolver, contracts
 │   ├── orchestrator/       # Flagship muse_context fusion, bidirectional lookups, MCP profiles
 │   ├── paths.ts            # Strict path containment (resolveWithin) & intentional-external opt-out (resolveExternal)
@@ -81,10 +81,10 @@ musememory/
 │   ├── types.ts            # Core TypeScript interfaces & enums
 │   ├── ui.ts               # Full Web Observability Studio & 3D knowledge graph
 │   ├── user.ts             # Persona & working preference engine
-│   ├── agents/             # Workstation coding agents detection & registry (80+ platforms)
-│   └── migrator/           # Multi-provider auto-detection & migration engine (24+ formats)
-├── test/                   # Automated Bun test suites (498 tests across 81 suites)
-├── package.json            # Package metadata & dependencies (v2.0.0)
+│   ├── agents/             # Workstation coding agents detection & registry (52 platforms)
+│   └── migrator/           # Multi-provider auto-detection & migration engine (30 providers)
+├── test/                   # Automated Bun test suites (1592 tests across 252 suites)
+├── package.json            # Package metadata & dependencies (v3.4.0)
 └── dist/                   # Compiled Node/Bun bundle
 ```
 

package/.agents/context/current.md

@@ -1,7 +1,160 @@
 # 📍 Current Shipped State & Active Invariants
 
-> **Version**: `v3.1.3`  
-> **Status**: Wave 29 Landed — Sovereign LifeOS Migration & Complete Decommissioning (T-444 to T-448).
+> **Version**: `v3.4.0` (post-release `dev`)  
+> **Status**: Wave 29 Landed — Sovereign Ecosystem Migration & Complete Decommissioning (T-444 to T-448).
+
+---
+
+## 🧾 Session — Changelog Repair & D5 Write Race (2026-11-10)
+
+**v3.3.0 shipped with no changelog section.** The tag and the npm publish
+both went out while all 29 release bullets were still under `[Unreleased]`,
+so the published release notes omitted the entire release; the section also
+carried two separate `### Added` headings. Promoted the 29 bullets verbatim
+into a dated `[3.3.0]` section (verified byte-identical against
+`git show v3.3.0:CHANGELOG.md`), rebuilt `[Unreleased]` from only the 12
+post-tag bullets, and added the five changes from this session. There is no
+CI check for changelog structure — this was only caught by reading it.
+
+**D5 — playbook writes lost data across processes.** The debt said "random
+temp suffix + fsync". Implementing exactly that did **not** fix the bug: 8
+concurrent processes still persisted only 5. The temp-name fix prevents
+staging collisions, but `savePlaybook` and `recordPlaybookOutcome` are
+read-modify-write, so both processes read the same list and the later rename
+drops the earlier write. Wrapping both in the existing `withStoreLock`
+fixed it — 8/8 persist. Outcome recording had the same race, silently
+corrupting the success/failure ratio ranking depends on.
+
+**Lesson recorded:** the first regression test used an in-process
+`Promise.all` race and passed with *and* without the fix, because the write
+path is synchronous and can never interleave in one process. It was
+replaced with a real multi-process test; both guards now fail against the
+old implementation (6 expected, 3 received).
+
+Gate: `bun test` 1592/1592 pass across 252 files; `bunx tsc --noEmit`
+clean; `check:landed` → **LANDED** at `9270d97`.
+
+---
+
+## 🧾 Session — Stack Landing, W26-5 CTE Bound & PR Queue Cleanup (2026-11-10)
+
+**CI gap on stacked PRs (#390).** The `pull_request` trigger listed only
+`main`/`master`/`dev`. Retargeting stacked PRs onto their parent branches
+therefore left three of four layers with **no CI at all** — they reported
+"no checks reported" and looked fine to a reviewer. Widened the filter to
+`["**"]`. Note for future stack work: GitHub reads `pull_request`
+workflows from the **base** branch, so the fix has to reach the base first.
+Retargeting alone emits no event; a close/reopen is what re-triggers CI.
+
+**W26-5 — recursive CTE bound.** `maxNodes` was clamped and passed to SQL,
+but `LIMIT` sat only on the outer `SELECT`, so SQLite materialised the whole
+reachable frontier before projecting. `preflightContext` calls this path on
+every file open. A `LIMIT` inside the recursive member took a 3000-node
+dense graph from **269.8ms to 9.7ms** (27x) with identical output. The
+regression test asserts a 50ms ceiling and was verified to *fail* without
+the fix (177ms) — a guard that only passes is not a guard.
+
+**W26-2 follow-up.** The over-budget message printed the pack total beside
+the word "over" — "192 tokens (over budget: 192 tokens)" reads as "over by
+192" when the overage was 42. Now "(over the 150 budget)". Caught by
+running the CLI, not by tests, which asserted engine fields but never the
+printed line.
+
+**PR queue cleaned.** Five spec PRs (#52–#56) were closed as superseded:
+their content already lives in `roadmap.md` in more detail, they targeted
+`main` against the `dev` convention, and each added a file to
+`.agents/plans/` — gitignored by ADR-008, so merging would have re-tracked
+a path the repo deliberately excluded.
+
+Gate: `bun test` 1590/1590 pass across 252 files; `bunx tsc --noEmit`
+clean; 162 doc links resolve; `check:landed` → **LANDED** at `df9e43f`.
+
+---
+
+## 🧾 Session — Agent Context Sync & AI-Readiness Closeout (2026-11-10)
+
+**Context drift repaired.** Agent-facing docs had fallen behind the engine:
+`AGENTS.md` pinned `v3.0.2` and `.agents/context/index.md` pinned `v3.1.0`
+while `package.json` is `3.3.0`. `architecture.md` still advertised 69 MCP
+tools, 80+ agent platforms, 24+ migrator formats, 498 tests across 81 suites
+and `v2.0.0`; the generated `claims.json` says 126 tools, 52 platforms, 30
+providers and 1584 tests across 251 files. All corrected from the generated
+source of truth, not from prose.
+
+**Roadmap debt closed.** The status table was dated `v2.3.0` and still listed
+"no branch-aware memory" as missing; `memory branch/checkout/blame` (R23) and
+branch-isolated overlays (T-436) both shipped, so the row now records them.
+
+**Standards adopted.** Five canonical core standards that were never present
+here are now materialised and routed from `AGENTS.md`: `anti-patterns.md`,
+`boundary-governance.md`, `test-failure-audit.md`, `language-policy.md` and
+`visual-inspection.md`. Curated local standards (project-specific naming, the
+project's own Landing Gate, `memory scan-secrets`) were deliberately **not**
+overwritten — the canon differs from them by design, and the sync engine
+reports that divergence rather than clobbering it.
+
+**Audit 9/14 → 14/14.** Added the assets the readiness matrix was missing:
+`.env.example` documenting the 15 real `MUSE_*` variables, `.mcp.json` wiring
+this repo's own MCP server (smoke-tested over stdio), a root `llms.txt` index,
+a `.agents/artifacts/README.md` contract stub, and the six Wave-23 terrain maps
+generated by the repo's own `writeWorkspaceTerrain` (not hand-written), which
+pass the terrain audit with zero credential leaks and zero hardcoded paths.
+
+Gate: `bun test` 1584/1584 pass, `bunx tsc --noEmit` clean, 162 doc links
+resolve, `--lint-context` clean, `--check` reports zero context drift.
+
+---
+
+## 🧾 Session — Outcome Study & W27-7 Read-Only Surfaces (2026-11-09)
+
+**Outcome study (EVD-008).** Replaced the two-task pilot with a ten-task,
+condition-blind study. Both arms received identical instructions and were told
+to consult memory; the only variable was whether ten project conventions were
+seeded. Result: **10/10 with memory vs 6/10 without**, four discriminating tasks
+(duration units, UTC dates, slug length cap, CSV whitespace quoting), zero
+regressions, Fisher one-sided p = 0.043. EVD-005 is no longer quarantined — its
+statement was narrowed to convention recall rather than an unqualified
+productivity claim. Evidence: `docs/evidence/2026-11-outcome-study.md` and
+`.agents/artifacts/outcome-study-2026-11/`.
+
+**W27-7 read-only MCP surfaces.** Added three read-only tools with explicit
+scope: `muse_enclave_recall` (client required, every read audit-logged),
+`muse_overlay_status` (list or inspect branch overlays, never mutates), and
+`muse_attestation_verify` (inline public key only — no model-supplied file
+paths). `src/store/branch.ts` gained `listBranchOverlays`. Tool count 119 → 122;
+claims and CLI reference regenerated; `test/mcp_readonly_surfaces.test.ts`
+covers the happy paths, the fail-loud cases and the read-only guarantee.
+Still open from the Wave 27 debt table: W27-8 (whole-vault rewrite) and W27-9
+(overlay size caps).
+
+**Follow-up (same day) — W27-8/W27-9 closed, R29 started.** Enclave vaults now
+store one AES-256-GCM file per entry under `<client>/<division>/`; a write no
+longer rewrites the vault, legacy single-file vaults stay readable and migrate
+on the next write, and per-entry (1 MiB) plus per-vault (5000) caps fail loud.
+Branch overlays are capped at 1000 entries / 2 MiB. `muse_enclave_recall` now
+redacts credentials with `redactSecrets` before returning content. On the
+operations side, `audit.jsonl` rotates at 5 MiB into `audit-archive/` (five
+segments retained by default), pruning writes a chain anchor so
+`getAuditTrail`/`getLastAuditHash`/`verifyAuditChain` remain valid across the
+retained window, and `memory io [--json]` reports `/proc/self/io` counters plus
+SQLite/WAL/audit/observation bytes. R29 is now complete except for optional
+async batching: the YAML mirror can be skipped with `MUSE_YAML_MIRROR=0` /
+`openStore({ yamlMirror: false })` (SQLite stays authoritative), and the WAL is
+bounded by `wal_autocheckpoint = 512` and `journal_size_limit = 64 MiB`.
+
+**Wave 28 complete (`T-440`–`T-443`).** The outcome loop is wired both ways:
+`test_feedback.ts` parses bun/vitest/jest/cargo/pytest/go output, binds a real run
+to the applied memory IDs, and reinforces green runs or records a failed
+application plus a T-313 regression strike only for anchor-relevant memories.
+`cognition/sentry.ts` matches failures against BUG_PRONE_PATTERN / DO_NOT_USE /
+failed-approach memories and returns reproduction/avoidance/recommendation
+playbooks without mutating anything. `doctor/flaky.ts` reconstructs the run
+history and reports flaky tests (per command), failure bursts and dependency-file
+drift before a burst. `cognition/remediation.ts` turns recurring failures into
+playbook candidates and persists resolved ones (dry-run by default). Surfaces:
+`memory test-feedback`, `sentry`, `flaky`, `remediation`, `memory_test_feedback`,
+`muse_regression_sentry`, `muse_flaky_tests`, `muse_remediation_playbooks`
+(126 tools).
 
 ---
 
@@ -33,12 +186,12 @@ Restored `AGENTS.md` byte-exact from HEAD after an `updateagents` run from the u
 
 ## 🏛️ Wave 29 — Sovereign Ecosystem Migration & Decommissioning Complete (2026-10-26)
 
-All personal identity, telos, operational rules, and hardware terrain scanning migrated into sovereign `musememory` engines, native Vibeguard secret scanning, and de-LifeOSified agent governance. Legacy import tools have been permanently purged from source:
+All personal identity, telos, operational rules, and hardware terrain scanning migrated into sovereign `musememory` engines, native Vibeguard secret scanning, and decommissioned legacy agent governance. Legacy import tools have been permanently purged from source:
 
-- **Sovereign Persona & Store Ingestion (`T-444`)**: Identity, telos, and operational rules synthesized directly into sovereign `~/.memory/USER.md` persona and SQLite store. Legacy tool `import_lifeos.ts` permanently deleted from source.
+- **Sovereign Persona & Store Ingestion (`T-444`)**: Identity, telos, and operational rules synthesized directly into sovereign `~/.memory/USER.md` persona and SQLite store. The legacy import tool was permanently deleted from source.
 - **Native Vibeguard CLI (`T-445`)**: `memory scan-secrets [dir]` standalone zero-dependency secret scanner replacing legacy external scripts (`test/cli_secrets.test.ts`).
 - **Machine Terrain Collector (`T-446`)**: `memory terrain [--out file.md]` compiles deterministic hardware, toolchain, and local agent ecosystem maps into `system.md` (`test/vault_system.test.ts`).
-- **Global Governance De-LifeOSification (`T-447`)**: Removed all 6 legacy imports and paths from `~/.agents/AGENTS.md` and repository standards, binding truth directly to `~/.memory/USER.md` and `memory` MCP.
+- **Global Governance Decommissioning (`T-447`)**: Removed all 6 legacy imports and paths from `~/.agents/AGENTS.md` and repository standards, binding truth directly to `~/.memory/USER.md` and `memory` MCP.
 - **Sovereignty Decommissioning Gate (`T-448`)**: Enforces zero external legacy imports, confirms legacy tool removal, and passes 100% cleanly in `test/sovereignty_gate.test.ts`.
 - **Golden Sequence Documentation**: Published the complete 4-phase best-practice sequence for using Muse Memory with AI coding agents in `README.md`, `docs/agent-reference.md`, and `docs/cli-reference.md`.
 
@@ -276,7 +429,7 @@ on any blocking failure, so CI runs it. Report:
   - `T-144`: Trajectory fix-pair mining discovering negative memories (`AVOID: ...`) from *Error ➔ Remediation ➔ No Recurrence* observation sequences with affinity filters excluding non-mutating commands (`ls`, `cat`), dynamic line number normalization, Vibeguard secret scrubbing, and session volume cap of max 3 candidates (`test/fix_pair_mining.test.ts`, commit `cc4b5c5`).
   - `T-314`: Source-tagged recall hits wiring `.memory/HOT.md` candidates into retrieval ranking with sub-millisecond memoized `mtime` cache, origin tagging (`source_channels: ("hot" | "fts" | "vector" | "tree" | "distilled")[]`), relevance-gated `sourceBonus` (+0.20), and explainability tokens (`source:hot`, `source:fts`) (`test/source_tagging_frequency.test.ts`, commit `f9f6a93`).
   - `T-315`: LangMem logarithmic frequency rank curve ($0.25 \times \frac{\ln(1 + \text{effectiveCount})}{\ln(1 + 50)}$) replacing unbounded linear frequency scoring; weights verified memory applications $2\times$ over passive context loads; 7-day recency bonus (+0.05) capped at 0.30; outputs `freq:high` tokens (`test/source_tagging_frequency.test.ts`, commit `f3c8f1d`).
-- [x] **Wave 21: LifeOS Distillation, Mutation Safety & Knowledge Routing**:
+- [x] **Wave 21: Ecosystem Distillation, Mutation Safety & Knowledge Routing**:
   - `T-050`: Overnight & in-lock mutation safety guards (`verifyMutationSafety`, `MutationSafetyError`, `ESUSPECT_SHRINK`, `ESUSPECT_EROSION`, `allowDrastic`) preventing accidental bulk erasure or semantic erosion during compaction/governance (`test/mutation_guards.test.ts`, commit `439e31c`).
   - `T-177`: Pre-mutation 30-snapshot rolling ring buffer (`captureSnapshotRing`, `listSnapshotRing`, `restoreSnapshotRing`) with zero resident daemons in `.memory/snapshots/` before write transactions and compaction handoffs (`test/snapshot_ring.test.ts`, commit `76c58eb`).
   - `T-324`: Diff-preserving `CURRENT.md` checkpoint merge semantics (`parseCurrentConstraints`, `mergeCurrentConstraints`), preventing mid-session hand-added constraints from being overwritten (`test/current_compaction_guard.test.ts`, commit `f1a4756`).
@@ -474,4 +627,4 @@ Removed remaining competitor product names and external comparison links from we
 ## 1. Verified Shipped Reality
 - **2026-10-22**: T-321 per-project health gate scope
 - **2026-10-22**: T-320 contradiction engine calibration
-- **2026-10-20**: Wave 21: LifeOS Distillation, Mutation Safety & Knowledge Routing (T-050, T-177, T-324)
+- **2026-10-20**: Wave 21: Ecosystem Distillation, Mutation Safety & Knowledge Routing (T-050, T-177, T-324)

package/.agents/context/evidence-ledger.md

@@ -5,16 +5,16 @@
 | Category | Count |
 |---|---|
 | DECISION | 1 |
-| CLAIM | 5 |
+| CLAIM | 6 |
 | STATUS | 1 |
 
 | Status | Count |
 |---|---|
-| VERIFIED | 4 |
-| QUARANTINED | 1 |
-| ACTIVE | 1 |
+| VERIFIED | 6 |
+| ACTIVE | 2 |
+| QUARANTINED | 0 |
 
-**Health score**: 5 / 6 verifiable entries (83%) — one claim is quarantined, not contradicted.
+**Health score**: 6 / 6 claims verified (100%) — the previous quarantine on EVD-005 is resolved by EVD-008; the remaining two entries are an active fix record and an active decision.
 
 ---
 
@@ -59,12 +59,12 @@
 ---
 
 ### EVD-005 | 2026-11-09 | CLAIM
-**Statement**: Muse Memory improves coding-agent task success.
+**Statement** *(revised)*: On convention-sensitive coding tasks, an agent that can recall the project's stored conventions produces more correct implementations than the same agent without them.
 **Category**: `CLAIM`
-**Confidence**: `[INFER]` — plausible from retrieval quality, but no with/without task-outcome study exists.
-**Epistemology**: `[SPECULATIVE]`
-**Evidence**: none yet; retrieval evidence only (EVD-002/003).
-**Status**: `QUARANTINED` — a 2-task pilot (EVD-007) is suggestive, not sufficient.
+**Confidence**: `[EMPIRICAL]` — a ten-task condition-blind study (EVD-008) measured 10/10 with memory against 6/10 without it; four tasks discriminated and none regressed.
+**Epistemology**: `[EMPIRICAL]`
+**Evidence**: `docs/evidence/2026-11-outcome-study.md`; task prompts and hidden graders in `.agents/artifacts/outcome-study-2026-11/`.
+**Status**: `VERIFIED` — scoped to convention recall in this study. The original unqualified sentence ("improves coding-agent task success") is too broad for n=10 and has been narrowed rather than marked proven.
 
 ---
 
@@ -86,4 +86,14 @@
 **Confidence**: `[RAW]` — hidden Bun tests run against both fixture implementations.
 **Epistemology**: `[EMPIRICAL]`
 **Evidence**: `docs/evidence/2026-11-outcome-pilot.md`.
-**Status**: `VERIFIED` — pilot only (n=2, single run).
+**Status**: `VERIFIED` — pilot only (n=2, single run). Superseded as primary evidence by EVD-008.
+
+---
+
+### EVD-008 | 2026-11-09 | CLAIM
+**Statement**: In a ten-task condition-blind study, an agent told to consult memory passed 10/10 hidden tests when the conventions were seeded and 6/10 when they were not; the four discriminating tasks were duration units, UTC dates, the slug length cap and CSV whitespace quoting.
+**Category**: `CLAIM`
+**Confidence**: `[RAW]` — hidden Bun tests graded all twenty solutions mechanically under a non-UTC timezone; Fisher's exact one-sided p = 0.043.
+**Epistemology**: `[EMPIRICAL]`
+**Evidence**: `docs/evidence/2026-11-outcome-study.md`; `.agents/artifacts/outcome-study-2026-11/`.
+**Status**: `VERIFIED` — n=10, one run per arm, one model. Supports EVD-005 within that scope only.

package/.agents/context/index.md

@@ -13,7 +13,7 @@
 | [`product.md`](./product.md) | Product scope, the live MCP tool matrix, the 7 task profiles, the CLI surface, and the trust gates an agent must assume | Understanding user commands, capabilities, and MCP protocols |
 | [`architecture.md`](./architecture.md) | Dual-Scope storage (`.memory/`), SQLite DB, retrieval scoring formula, and file layout | System design, file locations, state machine, and data flow |
 | [`brand.md`](./brand.md) | Persona profiles (`USER.md` archetypes), voice, and presentation rules | Writing prompts, terminal output copy, and user documentation |
-| [`current.md`](./current.md) | Current shipped state (**v3.1.0**), test coverage, and active constraints | Before starting work on bugs, features, or refactors |
+| [`current.md`](./current.md) | Current shipped state (**v3.4.0**), test coverage, and active constraints | Before starting work on bugs, features, or refactors |
 | [`decisions.md`](./decisions.md) | Locked architectural decisions (zero-daemon, file-backed SQLite, atomic I/O) | Evaluating technical choices or refactors |
 | [`roadmap.md`](./roadmap.md) | Planned improvements, ecosystem expansions, and backlog | Scoping new capabilities |
 | [`ticket-plan.md`](./ticket-plan.md) | Detailed ticket plan for the next work waves | Planning a specific wave of work |