@rubytech/create-maxy-code 0.1.463 → 0.1.464

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@rubytech/create-maxy-code",
-  "version": "0.1.463",
+  "version": "0.1.464",
   "description": "Install Maxy — AI for Productive People",
   "bin": {
     "create-maxy-code": "./dist/index.js"

package/payload/platform/docs/superpowers/plans/2026-07-17-task-1736-retire-mcp-stderr-tee.md

@@ -0,0 +1,397 @@
+# Retire `mcp-stderr-tee` — Implementation Plan
+
+> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
+
+**Goal:** Delete the `mcp-stderr-tee` module, its 21 call sites, and every doc claim that a per-conversation MCP stream log exists — leaving `mcp-spawn-tee`'s per-session raw sink as the single, honest MCP-stderr surface.
+
+**Architecture:** This is removal work, not feature work. Each task asserts the current state (grep/build), makes the deletion, then asserts the new state. There is no new runtime behaviour to test; the invariants are grep counts and a clean build. The order is fixed: remove the importers before the module they import, or the build breaks mid-plan.
+
+**Tech Stack:** TypeScript (platform workspace at `maxy-code/platform/`), npm workspaces, Node ≥ 22.6.
+
+## Global Constraints
+
+- Work happens in worktree `worktree-task-1736-retire-mcp-stderr-tee`. All paths below are relative to its `maxy-code/` subdirectory unless absolute.
+- Node ≥ 22.6: `export PATH="$HOME/.nvm/versions/node/v22.22.0/bin:$PATH"` before any node/npm command.
+- Deps already installed at `maxy-code/platform/` via `npm install --ignore-scripts`.
+- **Collision guard (Task 5):** task 1737 owns `mcp-spawn-tee`'s stale marker names (`[mcp-spawn-tee-attached]`, `[mcp-spawn-tee-exit]`). Do not rename, restate, or fix them. Where a paragraph this plan edits also contains one, edit only the `initStderrTee`/`STREAM_LOG_PATH` clause and leave the stale name untouched for 1737.
+- **Success invariants (whole plan):** after all tasks, in the live tree (excluding `.tasks/archive/` and `docs/superpowers/`): `initStderrTee` → 0 hits, `mcp-stderr-tee` → 0 hits, `STREAM_LOG_PATH` → 0 hits.
+- Shipped-doc prose rule: the rewritten `plugins-guide.md` prose carries zero em-dashes (per repo doctrine for shipped markdown).
+
+---
+
+### Task 1: Remove the 21 `initStderrTee` call sites
+
+**Files:**
+- Modify: 20 plugin MCP entry files `platform/plugins/<name>/mcp/src/index.ts` for name in: admin, aeo, browser, connector, contacts, email, filesystem, graph-viewer, joblogic, memory, quickbooks, replicate, scheduling, storage-broker, telegram, url-get, voice-mirror, whatsapp, work, workflows
+- Modify: `platform/lib/graph-mcp/src/index.ts` (import line 31, call line 63 — NOT adjacent; leave `syncEmit` for Task 3)
+
+**Interfaces:**
+- Consumes: nothing.
+- Produces: nothing. This removes a call; no symbol is exported.
+
+**Non-uniform sites to handle by hand, not by a blind 2-line delete:**
+- `voice-mirror`: import is line 1, call is line 3. Line 2 is `import { lifelineTool } ...` — KEEP line 2.
+- `graph-mcp`: import line 31 (`import { initStderrTee } from "../../mcp-stderr-tee/dist/index.js";`), call line 63 (`initStderrTee(SERVER_NAME);`). Two separate deletions. Keep line 30 (`StringDecoder`, used at line 1001) and `SERVER_NAME` (used by `syncEmit` at line 83).
+- All other 19: import and call are adjacent (e.g. memory lines 1-2; whatsapp lines 8-9; admin 1-2; storage-broker 8-9; url-get 9-10; graph-viewer 7-8). Delete both lines.
+
+- [ ] **Step 1: Assert the starting count is 21**
+
+Run:
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code
+/usr/bin/grep -rl 'initStderrTee' platform/plugins/*/mcp/src/index.ts platform/lib/graph-mcp/src/index.ts | wc -l
+```
+Expected: `21`
+
+- [ ] **Step 2: Remove the import + call from each of the 20 plugin files**
+
+For the 19 adjacent sites, delete the `import { initStderrTee } ...` line and the `initStderrTee("<name>");` line. For `voice-mirror`, delete only the import (line 1) and the call (line 3), preserving the `lifelineTool` import. Use Edit per file with the exact two strings; do not delete by line number blindly (line numbers shift as you go).
+
+- [ ] **Step 3: Remove the import + call from graph-mcp**
+
+Delete `import { initStderrTee } from "../../mcp-stderr-tee/dist/index.js";` (line 31) and `initStderrTee(SERVER_NAME);` (line 63). Leave the surrounding `StringDecoder` import and `SERVER_NAME` const in place.
+
+- [ ] **Step 4: Assert zero call sites remain**
+
+Run:
+```bash
+/usr/bin/grep -rn 'initStderrTee' platform/plugins platform/lib | /usr/bin/grep -v node_modules | /usr/bin/grep -v '/dist/'
+```
+Expected: no output.
+
+- [ ] **Step 5: Build the lib dependency dists (needed for plugin typecheck), then typecheck the plugins**
+
+The plugin MCPs import sibling lib dists (`mcp-lifeline` etc.). Build libs first, EXCLUDING the soon-to-be-deleted `mcp-stderr-tee` (its dist may still exist from install; that's fine for this step — it just won't be imported anymore).
+
+Run:
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code/platform
+export PATH="$HOME/.nvm/versions/node/v22.22.0/bin:$PATH"
+npm run build:lib
+NODE_OPTIONS='--max-old-space-size=8192' npx tsc -b plugins/*/mcp/tsconfig.json lib/graph-mcp/tsconfig.json
+```
+Expected: exit 0, no errors. A wall of `TS7031 implicitly any` here means the lib dists did not build — fix setup, do not touch code (per project memory).
+
+- [ ] **Step 6: Commit**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee
+git add -A
+git commit -m "$(printf 'refactor: remove 21 initStderrTee call sites\n\nThe call is a no-op in production (STREAM_LOG_PATH is never set).\nmcp-spawn-tee already captures the same stderr per session.\n\nSession: 71944c42-a91c-452a-96d2-b7bbcfe888fb')"
+```
+
+---
+
+### Task 2: Delete the module and its build entries
+
+**Files:**
+- Delete: `platform/lib/mcp-stderr-tee/` (whole directory)
+- Modify: `platform/package.json` lines 9 (`build`) and 10 (`build:lib`) — remove ` && tsc -p lib/mcp-stderr-tee/tsconfig.json` from both
+
+**Interfaces:**
+- Consumes: nothing (Task 1 removed all importers).
+- Produces: nothing.
+
+- [ ] **Step 1: Assert no importer remains before deleting**
+
+Run:
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code
+/usr/bin/grep -rl 'mcp-stderr-tee/dist' platform premium-plugins | /usr/bin/grep -v node_modules | /usr/bin/grep -v '/dist/' | /usr/bin/grep -v '.tasks/'
+```
+Expected: no output. (If any hit, return to Task 1 — deleting the module now would break that importer.)
+
+- [ ] **Step 2: Delete the module directory**
+
+```bash
+git rm -r platform/lib/mcp-stderr-tee
+```
+
+- [ ] **Step 3: Remove both tsconfig entries from package.json**
+
+Edit `platform/package.json`: in the `build` script (line 9) and the `build:lib` script (line 10), delete the exact substring ` && tsc -p lib/mcp-stderr-tee/tsconfig.json` (it appears once per line, immediately after `lib/models/tsconfig.json`).
+
+- [ ] **Step 4: Verify build:lib still runs clean without the entry**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code/platform
+export PATH="$HOME/.nvm/versions/node/v22.22.0/bin:$PATH"
+npm run build:lib
+```
+Expected: exit 0. No reference to `mcp-stderr-tee`.
+
+- [ ] **Step 5: Commit**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee
+git add -A
+git commit -m "$(printf 'refactor: delete mcp-stderr-tee module and its build entries\n\nSession: 71944c42-a91c-452a-96d2-b7bbcfe888fb')"
+```
+
+---
+
+### Task 3: Neutralize the remaining `STREAM_LOG_PATH` readers
+
+Two readers survive Task 1: graph-mcp's `syncEmit` writes `[mcp:graph]` to `process.env.STREAM_LOG_PATH` on exit paths, and the memory boot-smoke harness sets it. Both point at the same never-produced destination. The `syncEmit` stderr fallback already reaches the `mcp-spawn-tee` sink, so removing the dead branch loses nothing.
+
+**Files:**
+- Modify: `platform/lib/graph-mcp/src/index.ts` — `syncEmit` (approx lines 66-92): remove the `STREAM_LOG_PATH` branch and its now-unused locals
+- Modify: `platform/plugins/memory/mcp/scripts/boot-smoke.sh:46` — remove the `STREAM_LOG_PATH="$LOG_DIR/stream.log" \` line
+
+**Interfaces:**
+- Consumes: nothing.
+- Produces: `syncEmit(line: string): void` — unchanged signature, still called at all 5 exit paths; only its body shrinks.
+
+- [ ] **Step 1: Assert the three graph-mcp hits and one boot-smoke hit exist**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code
+/usr/bin/grep -n 'STREAM_LOG_PATH\|streamLogPath' platform/lib/graph-mcp/src/index.ts
+/usr/bin/grep -n 'STREAM_LOG_PATH' platform/plugins/memory/mcp/scripts/boot-smoke.sh
+```
+Expected: 3 graph-mcp lines (80, 81, 83), 1 boot-smoke line (46).
+
+- [ ] **Step 2: Rewrite `syncEmit` to drop the dead branch**
+
+Replace the current body (which reads `const iso`, `const streamLogPath = process.env.STREAM_LOG_PATH`, the `if (streamLogPath) { try { appendFileSync(...) } catch {} }` block, then the stderr write) with:
+
+```typescript
+function syncEmit(line: string): void {
+  const msg = `[graph-mcp] ${line}`;
+  // The mcp-spawn-tee shim (PLUGIN.md spawns this entry under it) installs an
+  // appendFileSync-backed stderr writer, so this line reaches the raw sink
+  // synchronously and survives an immediate process.exit().
+  try {
+    process.stderr.write(`${msg}\n`);
+  } catch {
+    /* stderr closed — nothing else to do */
+  }
+}
+```
+
+This drops the `iso` local and the `appendFileSync` call. Then remove `appendFileSync` from the `node:fs` import at line 28 (it has no other use — the file uses `accessSync`, `constants`, `readFileSync`, `statSync`). Verify before removing:
+
+```bash
+/usr/bin/grep -n 'appendFileSync' platform/lib/graph-mcp/src/index.ts
+```
+Expected after the syncEmit rewrite: only the line-28 import mention remains, so remove it from the import list.
+
+- [ ] **Step 3: Remove the boot-smoke assignment**
+
+Delete the line `  STREAM_LOG_PATH="$LOG_DIR/stream.log" \` from `platform/plugins/memory/mcp/scripts/boot-smoke.sh` (line 46, inside the `env` block). The surrounding `LOG_DIR=` and `NEO4J_URI=` lines stay.
+
+- [ ] **Step 4: Assert zero `STREAM_LOG_PATH` hits in the live tree**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code
+/usr/bin/grep -rn 'STREAM_LOG_PATH' platform premium-plugins packages | /usr/bin/grep -v node_modules | /usr/bin/grep -v '/dist/' | /usr/bin/grep -v '.tasks/' | /usr/bin/grep -v 'docs/superpowers/'
+```
+Expected: no output.
+
+- [ ] **Step 5: Typecheck graph-mcp and confirm boot-smoke still parses**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code/platform
+export PATH="$HOME/.nvm/versions/node/v22.22.0/bin:$PATH"
+npx tsc -p lib/graph-mcp/tsconfig.json
+bash -n plugins/memory/mcp/scripts/boot-smoke.sh
+```
+Expected: both exit 0.
+
+- [ ] **Step 6: Commit**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee
+git add -A
+git commit -m "$(printf 'refactor: drop the two remaining dead STREAM_LOG_PATH readers\n\ngraph-mcp syncEmit and the memory boot-smoke harness both wrote to a\nvariable no spawner sets. syncEmit keeps its stderr write, which the\nmcp-spawn-tee shim already captures to the raw sink.\n\nSession: 71944c42-a91c-452a-96d2-b7bbcfe888fb')"
+```
+
+---
+
+### Task 4: Strip the three comment-only references
+
+**Files:**
+- Modify: `platform/services/claude-session-manager/src/rc-daemon.ts:460` — comment naming `mcp-stderr-tee`
+- Modify: `platform/scripts/logs-rotate.sh:182` — comment naming `mcp-stderr-tee`
+- Modify: `packages/create-maxy-code/scripts/bundle.js:563` — the lib-rewrite comment example
+
+**Interfaces:** none — comments only.
+
+- [ ] **Step 1: Read each comment in context**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code
+sed -n '458,462p' platform/services/claude-session-manager/src/rc-daemon.ts
+sed -n '180,184p' platform/scripts/logs-rotate.sh
+sed -n '561,565p' packages/create-maxy-code/scripts/bundle.js
+```
+
+- [ ] **Step 2: Edit each so the sentence stays true without the dead module name**
+
+- `rc-daemon.ts:460`: the comment says the sink is off `mcp-stderr-tee` "which no longer reads LOG_DIR". Reword to describe the current state without the retired module (the sink is `mcp-spawn-tee`'s `mcp-<name>-<SESSION_ID>.log`).
+- `logs-rotate.sh:182`: the comment references why an earlier draft could not purge `mcp-*` because `mcp-stderr-tee` held the file open. Reword to name `mcp-spawn-tee` (the current holder) or drop the historical clause.
+- `bundle.js:563`: change the example path in the comment from `mcp-stderr-tee` to a lib premium plugins actually import (e.g. `mcp-lifeline`). Verify the replacement is real:
+  ```bash
+  /usr/bin/grep -rl 'mcp-lifeline/dist' premium-plugins | /usr/bin/grep -v node_modules | head -1
+  ```
+  Expected: at least one premium plugin imports it.
+
+- [ ] **Step 3: Assert zero `mcp-stderr-tee` hits in the live tree**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code
+/usr/bin/grep -rn 'mcp-stderr-tee' platform premium-plugins packages | /usr/bin/grep -v node_modules | /usr/bin/grep -v '/dist/' | /usr/bin/grep -v '.tasks/' | /usr/bin/grep -v 'docs/superpowers/' | /usr/bin/grep -v 'plugins/docs/references/plugins-guide.md' | /usr/bin/grep -v 'platform-architecture/SKILL.md'
+```
+Expected: no output. (plugins-guide.md and SKILL.md are handled in Tasks 5-6.)
+
+- [ ] **Step 4: Commit**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee
+git add -A
+git commit -m "$(printf 'refactor: strip dead mcp-stderr-tee name from three comments\n\nSession: 71944c42-a91c-452a-96d2-b7bbcfe888fb')"
+```
+
+---
+
+### Task 5: Rewrite `plugins-guide.md` so it stops claiming a per-conversation stream log
+
+**Files:**
+- Modify: `platform/plugins/docs/references/plugins-guide.md` (the `## MCP Plugin Observability` section, ~lines 153-197)
+
+**Interfaces:** none — documentation.
+
+**Keep / remove / rewrite map** (verify line numbers against the file first — they shift as you edit):
+
+| Content | Action |
+|---|---|
+| 155 intro ("Every `console.error` line ... teed into the per-conversation stream log") | REWRITE: `console.error` from a plugin's MCP server is captured by the `mcp-spawn-tee` shim into the per-session raw file `mcp-<name>-<SESSION_ID>.log`. |
+| 157-161 Opt-in block (`import { initStderrTee }` + call) | REMOVE entirely — no opt-in; capture is automatic under the shim. |
+| 163 ("After this ... `mcp-your-plugin-name-{sessionId}.log` ... That is where you grep") | KEEP — true; it names the real sink. Merge with the rewritten intro. |
+| 166 ("`initStderrTee` is *intended* to also mirror ... no-op in production ... Tracked as 1736") | REMOVE — the whole reason to retire; do not leave a "tracked as 1736" pointer to a closed task. |
+| 168 Premium-plugins paragraph (ships `mcp-stderr-tee/{dist,package.json}`) | REMOVE — it documents shipping the deleted module. The generic bundler behaviour is not plugin-author guidance and lives in `bundle.js`. |
+| 170 ("How the tee decides ... platform sets `STREAM_LOG_PATH` on every spawn") | REMOVE — false claim. |
+| 174-177 Retrieve block: `logs-read {type:"system"}` → grep stream log | REMOVE the `type:"system"` bullet (returns nothing). KEEP the `type:"mcp"` bullet (real raw feed). |
+| 179 Tee-state markers (`[mcp-tee-attach/skip/detach]` land in stream log) | REMOVE — no such log. |
+| 181 Main-subprocess stderr (`[subproc-stderr]`) | REMOVE — describes the retired admin server. |
+| 185 (`initStderrTee` `createWriteStream` write-race paragraph) | REMOVE — about the deleted module. |
+| 187-191 "Two layers" — plugin-side sync-write + `mcp-spawn-tee` wrapper | KEEP, but in the wrapper paragraph edit only clauses that name `initStderrTee` (e.g. "module-load throws before `initStderrTee` runs" → "module-load throws before any plugin logging runs"). DO NOT touch `[mcp-spawn-tee-attached]` / SIGTERM-forwarding wording — that is 1737's collision surface. |
+
+- [ ] **Step 1: Re-read the section with current line numbers**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code
+sed -n '153,197p' platform/plugins/docs/references/plugins-guide.md
+```
+
+- [ ] **Step 2: Apply the keep/remove/rewrite map**
+
+Edit the section per the table. Prose carries zero em-dashes. The section should end shorter, asserting one surface: automatic capture by the shim into the per-session raw sink, retrievable via `logs-read {type:"mcp"}`, with the two-layer failure contract intact minus the `initStderrTee` clauses.
+
+- [ ] **Step 3: Assert the false surfaces are gone and 1737's names are untouched**
+
+```bash
+/usr/bin/grep -n 'initStderrTee\|STREAM_LOG_PATH\|subproc-stderr\|mcp-tee-attach\|claude-agent-stream\|logs-read { type: "system"' platform/plugins/docs/references/plugins-guide.md
+```
+Expected: no output.
+```bash
+/usr/bin/grep -c 'mcp-spawn-tee-attached' platform/plugins/docs/references/plugins-guide.md
+```
+Expected: `1` (unchanged — still 1737's to fix, not zero).
+
+- [ ] **Step 4: Commit**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee
+git add -A
+git commit -m "$(printf 'docs: plugins-guide no longer claims a per-conversation MCP stream log\n\nThe mcp-spawn-tee per-session raw sink is the single honest surface.\nLeaves 1737 stale marker names untouched.\n\nSession: 71944c42-a91c-452a-96d2-b7bbcfe888fb')"
+```
+
+---
+
+### Task 6: Regenerate the architecture skill and pass the no-drift gate
+
+**Files:**
+- Regenerate: `platform/plugins/admin/skills/platform-architecture/SKILL.md` (and any brand twins) via the assembler
+- Possibly regenerate: `docs/public/llms-full.*.txt` corpus via copy-docs
+
+**Interfaces:** none — generated artefact.
+
+- [ ] **Step 1: Refresh the doc corpus and reassemble the skill**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code
+export PATH="$HOME/.nvm/versions/node/v22.22.0/bin:$PATH"
+node scripts/copy-docs.mjs 2>&1 | tail -3
+node scripts/assemble-architecture-skill.mjs 2>&1 | tail -3
+```
+Expected: both complete without error. (If `copy-docs.mjs` needs args or a different path, read `check-architecture-skill-no-drift.mjs` for the exact corpus path it expects and match it.)
+
+- [ ] **Step 2: Run the no-drift gate**
+
+```bash
+node platform/scripts/check-architecture-skill-no-drift.mjs
+```
+Expected: exit 0, no "out of sync" failure.
+
+- [ ] **Step 3: Confirm the regenerated skill dropped the false surface**
+
+```bash
+/usr/bin/grep -n 'initStderrTee\|STREAM_LOG_PATH\|subproc-stderr' platform/plugins/admin/skills/platform-architecture/SKILL.md
+```
+Expected: no output.
+
+- [ ] **Step 4: Commit**
+
+```bash
+git add -A
+git commit -m "$(printf 'chore: regenerate platform-architecture skill after plugins-guide edit\n\nSession: 71944c42-a91c-452a-96d2-b7bbcfe888fb')"
+```
+
+---
+
+### Task 7: Correct the reasoning in the task file and LANES
+
+The archive MOVE happens in the sprint's Land phase (6c). This task fixes the CONTENT — both records currently argue "retire because 287/288 dismantled the consumers," which is wrong. Do this before the archive so the archived file carries the correct reason.
+
+**Files:**
+- Modify: `.tasks/1736-nothing-sets-stream-log-path-so-the-per-conversation-mcp-stream-log-does-not-exist.md` (reasoning already partly corrected in an earlier session; verify it names the real reason: `mcp-spawn-tee` already serves this consumer per session)
+- Modify: `.tasks/LANES.md` row 1736 (same correction)
+
+**Interfaces:** none — tracker prose.
+
+- [ ] **Step 1: Verify current task-file reasoning names the real reason**
+
+```bash
+cd /Users/neo/getmaxy/maxy-code/.claude/worktrees/task-1736-retire-mcp-stderr-tee/maxy-code
+/usr/bin/grep -n 'mcp-spawn-tee already\|287\|288\|dismantled' .tasks/1736-*.md
+```
+The task file must state the retire reason as redundancy with `mcp-spawn-tee`, and must not present "287/288 retired the consumers" as the load-bearing argument. If the current text still leans on 287/288 as the reason, reword it. (Note: the reasoning correction may already be applied from a prior session; this step confirms rather than assumes.)
+
+- [ ] **Step 2: Confirm LANES row 1736 matches**
+
+```bash
+/usr/bin/grep -n '287 retired the script_stream\|mcp-spawn-tee already writes' .tasks/LANES.md
+```
+Ensure the row's DECIDED block frames the reason as redundancy, with 287 named only as what it actually did (retired `script_stream`), not as the reason to retire this.
+
+- [ ] **Step 3: Commit any corrections**
+
+```bash
+git add -A
+git commit -m "$(printf 'docs: correct the retire reasoning in task 1736 and its LANES row\n\nSession: 71944c42-a91c-452a-96d2-b7bbcfe888fb')" || echo "nothing to correct — already accurate"
+```
+
+---
+
+## Self-Review
+
+**Spec coverage:** module deletion (T2), 21 call sites (T1), build entries (T2), boot-smoke assignment (T3), graph-mcp second reader (T3 — an addition the spec implied via "zero STREAM_LOG_PATH hits" but did not enumerate), comment refs (T4), doc rewrite (T5), skill regen (T6), reasoning correction (T7). The `logs-read {type:"system"}` open question is handled inside T5 (the doc bullet is removed) with the tool-level bug filed separately if implementation finds it reads a producerless file. All spec sections map to a task.
+
+**Placeholder scan:** no TBD/TODO; every edit names exact files, strings, and expected grep output.
+
+**Type consistency:** the only signature touched is `syncEmit(line: string): void` (T3), unchanged. No cross-task symbol references.
+
+## Deviation from spec, flagged
+
+Task 3's graph-mcp `syncEmit` edit is the one place the change rewrites live failure-path logic rather than deleting a no-op. It is in scope (it is a second reader of the same dead `STREAM_LOG_PATH`, and the spec's "zero hits" criterion requires it), and it is behaviour-preserving (the stderr write already reaches the sink). Called out here so a reviewer sees it deliberately, not buried in a bulk deletion.

package/payload/platform/docs/superpowers/specs/2026-07-17-task-1736-retire-mcp-stderr-tee-design.md

@@ -0,0 +1,128 @@
+# Task 1736 — retire `mcp-stderr-tee`
+
+**Date:** 2026-07-17
+**Task:** [`.tasks/1736-*.md`](../../../../.tasks/1736-nothing-sets-stream-log-path-so-the-per-conversation-mcp-stream-log-does-not-exist.md)
+**Verdict:** retire the module, its call sites, and every doc claim that a per-conversation
+MCP stream log exists.
+
+## Why retire, precisely
+
+`mcp-stderr-tee` reads `STREAM_LOG_PATH` and mirrors MCP stderr into a per-conversation
+stream log. Nothing sets that variable, so `initStderrTee` takes its skip path on every
+production spawn and 21 call sites wire up a no-op.
+
+The reason to retire is **not** that the surface was deliberately retired elsewhere. It is
+that `mcp-spawn-tee` already serves the same consumer, strictly better:
+
+- It is a shim Claude Code runs *in place of* each server's entry, so it replaces
+  `process.stderr.write` before the real entry imports. It captures every stderr byte with
+  no cooperation from the plugin — including module-load throws that happen before
+  `initStderrTee` would ever run.
+- It writes `${LOG_DIR}/mcp-<name>-<SESSION_ID>.log`, keyed by session, which *is* the
+  conversation. Per-conversation correlation already exists.
+- It handles the sessionless spawn explicitly (`mcp-<name>-nosession.log`).
+- It uses `appendFileSync`, which survives an immediate `process.exit`. `mcp-stderr-tee`
+  used an async `createWriteStream`, which does not.
+
+The two modules were never complements. Task 1721's docstring records that they wrote the
+same file and "every line landed twice". 1721 removed the duplication and left the loser
+holding a destination nothing produces.
+
+## Correcting the record
+
+The task file and its LANES row currently argue "retire because Tasks 287 and 288
+dismantled the consumers." That is wrong and it is load-bearing: 287 retired the
+`script_stream` surface — shell scripts narrating phase lines into the chat UI. It never
+evaluated the MCP-stderr surface, whose consumer is an investigator grepping a file. That
+consumer still exists; it is simply already served by `mcp-spawn-tee`.
+
+Both records get the correct argument before the task archives. A wrong reason in an
+archive licenses the next removal that the reason does not actually cover.
+
+## What changes
+
+**Code.**
+
+- Delete `platform/lib/mcp-stderr-tee/`.
+- Delete its two `tsc -p lib/mcp-stderr-tee/tsconfig.json` entries in
+  `platform/package.json` (`build` and `build:lib`).
+- Remove 21 `initStderrTee` call sites — 20 plugin MCP entry files plus
+  `platform/lib/graph-mcp/src/index.ts`. Each loses one import line and one call line.
+  `graph-mcp` is inside the 21, not additional. Premium plugins have zero call sites.
+- Remove the now-dead `STREAM_LOG_PATH` assignment in
+  `platform/plugins/memory/mcp/scripts/boot-smoke.sh` — the last assignment in the tree once
+  the reader is gone.
+- Strip the dangling module name from two comment-only references
+  (`services/claude-session-manager/src/rc-daemon.ts`, `platform/scripts/logs-rotate.sh`).
+  Neither imports it.
+- `packages/create-maxy-code/scripts/bundle.js` names `mcp-stderr-tee` as the example in its
+  lib-rewrite comment. The mechanism is generic and collects whatever premium plugins import;
+  the example just becomes stale. Point it at a lib that is actually imported.
+
+**Docs.** `platform/plugins/docs/references/plugins-guide.md` loses every claim that a
+per-conversation MCP stream log exists:
+
+- the `initStderrTee` opt-in block;
+- "How the tee decides which file to write to" (asserts the platform sets `STREAM_LOG_PATH`
+  on every spawn — it does not);
+- the tee-state markers paragraph (`[mcp-tee-attach]` / `[mcp-tee-skip]` /
+  `[mcp-tee-detach]` "land in the stream log");
+- the `logs-read { type: "system", sessionId }` → grep `[mcp:<name>]` retrieval instruction;
+- the "Main-subprocess stderr" `[subproc-stderr]` paragraph, which describes the retired
+  admin server;
+- the `initStderrTee` `createWriteStream` write-race paragraph in the failure-path contract;
+- the premium-plugins paragraph's `mcp-stderr-tee` bundler example.
+
+What replaces them is short: a plugin author calls `console.error`; the `mcp-spawn-tee` shim
+captures every byte into `mcp-<name>-<SESSION_ID>.log`; that file is where you grep. No
+import, no call, no opt-in.
+
+`platform/plugins/admin/skills/platform-architecture/SKILL.md` is generated from that
+reference doc. Regenerate it; never hand-edit.
+
+## Scope boundaries
+
+**In scope:** the `STREAM_LOG_PATH` contract — the module, its 21 call sites, its build
+entries, and the doc claims asserting the surface works.
+
+**Out of scope:**
+
+- The raw sink. Task 1721 settled it and this removal does not touch it.
+- `mcp-spawn-tee`'s stale marker names (`[mcp-spawn-tee-attached]` / `[mcp-spawn-tee-exit]`,
+  retired by Task 989 in favour of `[mcp-helper] op=spawn|boot|exit`) and its stale
+  two-process description. Owned by [`.tasks/1737-*.md`](../../../../.tasks/1737-mcp-spawn-tee-docs-promise-lifecycle-markers-and-a-child-process-that-task-989-removed.md),
+  which is designed but unimplemented on branch `worktree-task-1737-mcp-spawn-tee-docs`.
+  This sprint must not restate those names when it points authors at the sink, or it
+  re-introduces the exact bug 1737 exists to fix.
+- Reviving `claude-agent-stream-<id>.log` or any per-conversation agent stream file. It has
+  no producer in the PTY architecture; building one is a different task with a different
+  premise.
+- The `script_stream` surface and its banned-token gate. Tasks 287/288 settled it.
+
+## Open question to resolve during implementation
+
+`logs-read { type: "system", sessionId }` is documented as returning the conversation's
+stream log for `[mcp:<name>]` grepping. The doc instruction goes either way — grepping
+`[mcp:<name>]` on that log returns nothing, so the sentence is false regardless of what the
+tool does internally.
+
+What is unresolved is the *tool*. If `logs-read type:"system"` reads a file with no producer,
+that is a live bug and not this task's doc defect. Establish which during implementation. If
+it is a bug, file a `.tasks/NNN-*.md` for it — deleting the doc sentence while leaving a tool
+that silently returns nothing would hide the same false-clean shape one layer down, which is
+the exact failure this task exists to close.
+
+## Verification
+
+- `npm run build` at `maxy-code/platform/` is clean after the tsconfig entries are removed.
+- The 21 affected MCP packages typecheck.
+- `initStderrTee` and `mcp-stderr-tee` return zero hits outside `.tasks/archive/` and
+  `docs/superpowers/specs/`.
+- The architecture-skill regeneration gate passes with the regenerated `SKILL.md` committed.
+- `STREAM_LOG_PATH` returns zero hits in the live tree.
+
+## Deployment
+
+Every one of the 21 call sites ships in the bundled `create-maxy-code` payload, so this
+requires an installer publish to reach the fleet. Per the sprint gating rule, publish awaits
+an explicit operator command.

package/payload/platform/lib/graph-mcp/dist/index.js

@@ -29,7 +29,6 @@ const node_child_process_1 = require("node:child_process");
 const node_fs_1 = require("node:fs");
 const node_path_1 = require("node:path");
 const node_string_decoder_1 = require("node:string_decoder");
-const index_js_1 = require("../../mcp-stderr-tee/dist/index.js");
 const cypher_validate_js_1 = require("./cypher-validate.js");
 const audit_js_1 = require("../../graph-write/dist/audit.js");
 const schema_cache_js_1 = require("./schema-cache.js");
@@ -39,31 +38,13 @@ const cypher_shim_read_js_1 = require("./cypher-shim-read.js");
 const node_crypto_1 = require("node:crypto");
 const SERVER_NAME = "graph";
 const UPSTREAM_PACKAGE = "mcp-neo4j-cypher@0.6.0";
-(0, index_js_1.initStderrTee)(SERVER_NAME);
-// Sync-write failure line to the per-conversation stream log. Used only on exit
-// paths (missing password, uvx unresolvable, upstream spawn/exit error) where an
-// async createWriteStream buffer would not flush before process.exit() outraces
-// it. `appendFileSync` returns after the kernel has queued the write, so the line
-// survives an immediate exit. Each destination is wrapped independently — an
-// unwritable destination must not mask the primary failure.
-//
-// Task 1721 removed this function's own raw-file write. The final
-// process.stderr.write goes through the writer the mcp-spawn-tee shim installed
-// (PLUGIN.md spawns this entry under it), and that shim's appendSafe is itself
-// an appendFileSync to the raw sink — so exit-path lines still reach disk
-// synchronously, via one writer instead of two.
+// Emit a failure line on exit paths (missing password, uvx unresolvable,
+// upstream spawn/exit error). The mcp-spawn-tee shim spawns this entry (see
+// PLUGIN.md) and installs an appendFileSync-backed stderr writer, so this
+// process.stderr.write reaches the raw sink synchronously and survives an
+// immediate process.exit() — an async buffer would not flush in time.
 function syncEmit(line) {
     const msg = `[graph-mcp] ${line}`;
-    const iso = new Date().toISOString();
-    const streamLogPath = process.env.STREAM_LOG_PATH;
-    if (streamLogPath) {
-        try {
-            (0, node_fs_1.appendFileSync)(streamLogPath, `[${iso}] [mcp:${SERVER_NAME}] ${msg}\n`);
-        }
-        catch {
-            /* unwritable destination — preserve primary failure */
-        }
-    }
     try {
         process.stderr.write(`${msg}\n`);
     }