@mmerterden/multi-agent-pipeline 16.14.0 → 16.15.0

package/CHANGELOG.md

@@ -16,6 +16,19 @@ Internal file-layout changes that don't affect the slash-command surface are sti
 
 ## [Unreleased]
 
+## [16.15.0] - 2026-08-30
+
+### Fixed
+
+- **Usage reporting only ever reported runs that finished.** The reporter was called from Phase 7, and Phase 7 is the phase a run is least likely to reach: across the whole local log history, exactly one run got there. So the admin dashboard read zero of everything while the table held real rows, and the question it exists to answer - who is using this - had no answer. Phase 0's exit gate now reports the run as started, on the same upsert key, so Phase 7 replaces that row with the final state and nothing is double-counted. The read side of the same bug is fixed in the dashboard: its rollup folded only `terminal` rows, and an unfinished run was scored `fail`.
+
+
+### Changed
+
+- **`/multi-agent:help` now covers the plugins and the toolkit MCP.** It named both in passing - a `refactor` line, a `sync` line, a `design-check` footnote - and never said what they are or how to call one. The catalog gained a Plugins & tools block in both languages: how a stack skill is invoked, which plugins are always on, and the MCP's 83 tools grouped by what they drive (device control, web, design comparison, store and accessibility audits), with the note that its registration survives uninstall. Paid for by compressing the `design-check` entry, which was restating its own contract in fifteen lines: 353 tokens reclaimed before a byte was added, so the file sits within its 8500 ceiling. The tool count that went in with it was wrong and is now absent: `rules/outside-the-pipeline.md` said 83 while the server serves 84, and the pipeline cannot check either number because it does not vendor the toolkit. Both places now say 80+, which stays true; the exact count lives in the toolkit's own README, where a gate boots the server and compares `tools/list` against it.
+
+- **ADR-0010 now says where its numbers come from.** The coverage and token figures were published with no path to the harness that produced them, because the harness names a private codebase and cannot ship. It lives beside the graph it measured, under `~/.claude/knowledge/<project>/gate/`, and the ADR points at it: a number nobody can re-derive is an assertion, not a measurement.
+
 ## [16.14.0] - 2026-08-28
 
 ### Added

package/docs/adr/0010-own-code-graph.md

@@ -96,6 +96,12 @@ domain words, the graph scored 63.3% against 32.0% at 10,937 tokens against
 24,983. The value is in narrowing an open-ended search, not in replacing a
 grep for a name you already know.
 
+The harness, the ten graded questions and the raw per-question result live
+beside the graph they measured, under `~/.claude/knowledge/<project>/gate/`.
+They stay there rather than in this repo because they name the private
+codebase's files and symbols, and a number nobody can re-derive is an assertion
+rather than a measurement: `node run-gate.mjs` reproduces the table above.
+
 That measurement covers the context each strategy assembles, not the quality
 of an answer written from it. graphify's numbers measure the latter. The two
 are not directly comparable and this ADR does not claim they are.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mmerterden/multi-agent-pipeline",
-  "version": "16.14.0",
+  "version": "16.15.0",
   "description": "8-phase AI development pipeline with full orchestration on Claude Code, Copilot CLI and Codex CLI. Analysis, planning, TDD, CLI-aware parallel review with consensus surfacing + Fable triage, default-FAIL evidence gates, secret + intent guards, per-phase cost ledger, persistent learnings memory, wiki generation, commit automation. Token-preserving uninstall.",
   "type": "module",
   "main": "index.js",

package/pipeline/commands/multi-agent/help/SKILL.md

@@ -147,8 +147,17 @@ Setup & Maintenance:
 
 ------------------------------------------------------------
 
-Stack plugin skills: call directly as ai-<stack>-toolkit:<skill>. Catalog:
-  ai-<stack>-toolkit:help. Active plugins per repo: /multi-agent:stack.
+Plugins & tools (called directly, no wrappers):
+
+  Stack plugins    ai-<stack>-toolkit:<skill>, e.g. ai-ios-toolkit:create-component
+                   Catalog: ai-<stack>-toolkit:help   Enable per repo: /multi-agent:stack
+                   common + analyst are always on; the rest follow the repo's stack.
+  Toolkit MCP      multi-agent-toolkit  -  80+ tools for a RUNNING app, no pipeline run needed:
+                   ios_* / android_* (launch, tap, screenshot, ui tree, logs, crashes,
+                   locale, dark mode), web_*, design_* (Design Check above), and the
+                   audits (ios_app_store_audit, ios_testflight_validate, android_apk_audit,
+                   *_accessibility_audit). Use them instead of guessing about on-screen
+                   state. Its registration survives uninstall. Not registered = silent no-op.
 
 ------------------------------------------------------------
 
@@ -202,24 +211,17 @@ Manual Test (Phase 5 standalone  -  Xcode hint flow):
   /multi-agent:manual-test           Checkout task branch, print Xcode/SourceTree hints,
   /multi-agent:manual-test #N        wait for your "ok" / "fix: ..." verdict.
 
-Design Check (mock-mode vs Figma, local-only  -  powered by the toolkit MCP design_* tools):
-
-  /multi-agent:design-check                Whole-module design audit: pick repo+module, gate on mock support,
-                                           enumerate every state driver (launch args, scenario cases, scenario codes,
-                                           fixtures, deep links) into a countable target set, build Debug in a worktree,
-                                           drive each target in mock mode via idb, compare each vs its Figma frame
-                                           (padding/width/spacing/font/color/component/localization), and export a
-                                           report (HTML+PDF, ~/DesignChecks).
-  /multi-agent:design-check <screen>       Scope to one screen's targets (e.g. boarding-pass).
-  /multi-agent:design-check <screen@var>   Scope to a single variant (e.g. boarding-pass@expired).
-  /multi-agent:design-check <figma-url...>   Scope to the frame(s) those URLs name.
-  /multi-agent:design-check --resume       Audit the unaudited remainder of the last run for this repo+module.
-  # COVERAGE GATE: every target is either audited or skipped WITH a concrete reason. Anything else fails the
-  # gate and the run is reported INCOMPLETE with the missing target ids  -  a partial audit never reads as clean.
-  # "Requires a scenario / prefix / launch-arg" is not a reason; reaching those states is the run's job.
-  # When a live screen has no confident Figma match, it SHOWS you the screenshot and asks for the node-id.
-
-  Drives the toolkit design_* tools (+ ios_* / android_* device control).
+Design Check (mock-mode vs Figma, local-only):
+
+  /multi-agent:design-check                Whole-module audit: enumerate every state driver into a countable
+                                           target set, build Debug in a worktree, drive each in mock mode,
+                                           diff against its Figma frame, export HTML+PDF to ~/DesignChecks.
+  /multi-agent:design-check <screen>       Scope to one screen (e.g. boarding-pass).
+  /multi-agent:design-check <screen@var>   Scope to one variant (e.g. boarding-pass@expired).
+  /multi-agent:design-check <figma-url...> Scope to the frame(s) those URLs name.
+  /multi-agent:design-check --resume       Audit what the last run for this repo+module left unaudited.
+  # COVERAGE GATE: every target is audited or skipped WITH a concrete reason; anything else reports
+  # INCOMPLETE with the missing ids. "Needs a scenario/launch-arg" is not a reason, reaching it is the job.
 
 ------------------------------------------------------------
 
@@ -421,9 +423,17 @@ Setup & Maintenance:
 
 ------------------------------------------------------------
 
-Stack plugin skill'leri: doğrudan ai-<stack>-toolkit:<skill> olarak çağrılır
-  (örn. ai-ios-toolkit:create-component). Lokal kopya ve sarmalayıcı yok. Katalog:
-  ai-<stack>-toolkit:help. Aktif plugin'ler repo bazında: /multi-agent:stack.
+Plugin'ler ve tool'lar (doğrudan çağrılır, sarmalayıcı yok):
+
+  Stack plugin     ai-<stack>-toolkit:<skill>, örn. ai-ios-toolkit:create-component
+                   Katalog: ai-<stack>-toolkit:help   Repo bazında aç: /multi-agent:stack
+                   common + analyst hep açık; gerisi repo'nun stack'ini izler.
+  Toolkit MCP      multi-agent-toolkit  -  ÇALIŞAN uygulama için 80+ tool, pipeline koşusu
+                   gerektirmez: ios_* / android_* (başlat, tap, screenshot, ui ağacı, log,
+                   crash, locale, koyu mod), web_*, design_* (yukarıdaki Design Check), ve
+                   denetimler (ios_app_store_audit, ios_testflight_validate,
+                   android_apk_audit, *_accessibility_audit). Ekrandaki durumu tahmin etmek
+                   yerine bunları kullan. Kaydı uninstall'dan sağ çıkar. Kayıtlı değilse sessizce devre dışı.
 
 ------------------------------------------------------------
 
@@ -477,24 +487,17 @@ Manuel Test (Phase 5 standalone  -  Xcode hint akışı):
   /multi-agent:manual-test           Task branch'ine checkout, Xcode/SourceTree hint basar,
   /multi-agent:manual-test #N        "ok" / "fix: ..." yanıtını bekler.
 
-Design Check (mock-mod vs Figma, yalnızca lokal  -  multi-agent-toolkit-mcp design_* tool'larıyla):
+Design Check (mock-mod vs Figma, yalnızca lokal):
 
-  /multi-agent:design-check                Tüm modül tasarım denetimi: repo+modül seç, mock desteğini geçitle,
-                                           her state sürücüsünü (launch arg, senaryo case'leri, senaryo kodları,
-                                           fixture'lar, deep link'ler) sayılabilir bir hedef kümesine çıkar,
-                                           worktree'de Debug derle, her hedefi mock modda idb ile sür, her birini
-                                           Figma frame'iyle karşılaştır (padding/genişlik/spacing/font/renk/
-                                           component/localization), rapor çıkar (HTML+PDF, ~/DesignChecks).
-  /multi-agent:design-check <ekran>        Tek ekranın hedeflerine kapsa (örn. boarding-pass).
+  /multi-agent:design-check                Tüm modül denetimi: her state sürücüsünü sayılabilir hedef kümesine
+                                           çıkar, worktree'de Debug derle, her hedefi mock modda sür, Figma
+                                           frame'iyle diff'le, HTML+PDF raporu ~/DesignChecks'e yaz.
+  /multi-agent:design-check <ekran>        Tek ekrana kapsa (örn. boarding-pass).
   /multi-agent:design-check <ekran@varyant> Tek varyanta kapsa (örn. boarding-pass@expired).
-  /multi-agent:design-check <figma-url...>   URL'lerin işaret ettiği frame'lere kapsa.
-  /multi-agent:design-check --resume       Bu repo+modül için son koşunun denetlenmemiş kalanını denetle.
-  # KAPSAM GEÇİDİ: her hedef ya denetlenir ya da SOMUT bir gerekçeyle atlanır. Başka her durum geçidi
-  # düşürür ve koşu eksik hedef id'leriyle EKSİK raporlanır  -  kısmi denetim asla temiz görünmez.
-  # "Senaryo / prefix / launch-arg gerektirir" gerekçe değildir; o state'lere ulaşmak koşunun işidir.
-  # Canlı ekranın Figma karşılığı kesin bulunamazsa SANA screenshot'ı gösterip node-id sorar.
-
-  multi-agent-toolkit-mcp design_* tool takımını sürer (+ ios_* / android_* cihaz kontrolü).
+  /multi-agent:design-check <figma-url...> URL'lerin işaret ettiği frame'lere kapsa.
+  /multi-agent:design-check --resume       Son koşunun denetlenmemiş kalanını denetle.
+  # KAPSAM GEÇİDİ: her hedef ya denetlenir ya da SOMUT gerekçeyle atlanır; başka her durum eksik hedef
+  # id'leriyle EKSİK raporlanır. "Senaryo/launch-arg gerektirir" gerekçe değil, oraya ulaşmak koşunun işi.
 
 ------------------------------------------------------------
 

package/pipeline/multi-agent-refs/phases/phase-0-init.md

@@ -631,8 +631,12 @@ Phase 0 owns `agent-state.json`. Do not call
 
 ```bash
 node "$HOME/.claude/scripts/phase0-exit-gate.mjs" "$TASK_ID" --input "$ORIGINAL_INPUT"
+node "$HOME/.claude/scripts/usage-report.mjs" --task-id "$TASK_ID" >/dev/null 2>&1 || true
 ```
 
+The second line reports the run as started: reporting only from Phase 7 reported
+only runs that finish, and few do. Phase 7 upserts the same key over it.
+
 It asserts three things, each of which has failed silently in a real run:
 
 1. **`agent-state.json` exists.** A run once reported Phase 0 `completed` with only

package/pipeline/multi-agent-refs/phases/phase-7-report.md

@@ -196,7 +196,7 @@ $HOME/.claude/scripts/log-metric.sh "$TASK_ID" 7 task.completed \
   duration_ms=$TOTAL_DURATION
 ```
 
-**Operational reporting.** On by default since v16.8.0; `enabled: false` or `optOut: true` silences it. Sends command, duration, tokens, outcome - never a repo name or path. One POST per run, fire-and-forget, coarse run metadata only  -  no prompts, code, diffs, or absolute paths. The script no-ops when `usageLog.enabled` is not true or no token resolves, so the call is unconditional and never blocks the run.
+**Operational reporting.** On by default since v16.8.0; `enabled: false` or `optOut: true` silences it. Coarse run metadata only, never prompts, code, diffs, repo names or paths. Fire-and-forget: it no-ops without a token and never blocks the run, so the call is unconditional. Phase 0 already reported this run as started; this upserts the final state over it.
 
 ```bash
 node $HOME/.claude/scripts/usage-report.mjs --state "$STATE_FILE" >/dev/null 2>&1 || true

package/pipeline/rules/outside-the-pipeline.md

@@ -21,6 +21,6 @@ available. Read the effective `enabledPlugins` and load each enabled toolkit's
 `ai-common-toolkit` and `ai-analyst-toolkit` are on everywhere. Nothing enabled
 is a normal state.
 
-**multi-agent-toolkit MCP.** 83 tools for a running app: `ui-inspect`,
+**multi-agent-toolkit MCP.** 80+ tools for a running app: `ui-inspect`,
 `crash-logs`, `design-check`, `ios-app-store-audit`, `ios-testflight`. Use them
 instead of guessing about on-screen state. Not registered is a silent no-op.