@mmerterden/multi-agent-pipeline 19.0.0 → 19.1.0

package/CHANGELOG.md

@@ -14,6 +14,110 @@ Internal file-layout changes that don't affect the slash-command surface are sti
 
 ---
 
+## [19.1.0] - 2026-09-20
+
+Minor: the routing preference 19.0.0 shipped now reaches dispatch, and the
+toolkit's three new tool families are reflected everywhere the pipeline counts
+them.
+
+### Added
+
+- **`pipeline/lib/model-dispatch.sh` - routing that changes the answer.**
+  19.0.0 shipped `prefs.global.modelRouting` with a schema, four commands and a
+  status report, and nothing consulted it at dispatch time. A preference that is
+  written, validated and displayed but never read is worse than a missing one:
+  `route-status` said routing was on, the rules looked applied, and every call
+  went where it always went. Two call sites now ask: the subagent dispatch
+  contract in `skills/shared/core/multi-agent/SKILL.md`, and `bulk-read.sh`.
+
+  Precedence is `PHASE_MODEL_OVERRIDE` > a matching rule > persona
+  `preferredModel` > the global default. Routing sits below the per-dispatch
+  override on purpose: Phase 3 makes Reviewer 3 sonnet so the three reviewers
+  disagree, and a policy that could overrule that would turn a deliberate choice
+  into a suggestion.
+
+  The script never fails and never returns an empty rung - a router that can die
+  turns every call site into a place the run can die, for a feature that ships
+  disabled. Missing prefs, missing jq, unparseable JSON, an out-of-scope call
+  site and an unknown rung all return the caller's default, exit 0.
+
+  Two limits are enforced rather than documented. A rule preferring `fable`
+  falls past it while `modelFallback.fableEnabled` is false, so
+  `/multi-agent:model off` keeps meaning what it says. And a non-Anthropic rung
+  is refused for a subagent, because subagent dispatch belongs to the host - the
+  script says so on stderr instead of substituting an Anthropic rung and leaving
+  the user believing a rule worked that never could.
+- `cost-table.json` rungs declare a `provider`. Without it every rung looks
+  alike and the subagent limit above cannot be checked at all.
+- `smoke-model-dispatch.sh` (18 assertions). Half of them drive the router; the
+  other half assert the call sites invoke it, because a correct router nothing
+  calls is the same outage with better internals - which is exactly what 19.0.0
+  shipped.
+- **Six analysis Locked decisions gained a gate.** 13 (Section 4 scenarios are
+  Gherkin), 14 (a goal owes a paired non-goal), 15 (a new non-SVG asset owes a
+  rationale), 17 (Section 9 may not write "other errors" in place of a status
+  code), 18 (a screenshot is embedded, not linked to a host that outlives
+  nothing), and 21 (References is the last numbered section). Each checks a
+  shape the template prescribes and keys off a structure only a real document
+  carries, so a minimal fixture is skipped rather than failed. The Gate status
+  count moves 11 -> 18, and the 18 that stay prose now say WHY in three groups:
+  run behaviour no document records, evidence gathering that happened before
+  rendering, and judgement about meaning.
+- Nine anchors below Locked 25 in `smoke-locked-citations.sh`. The existing
+  anchors all sat above 25, which is where the v19 removal re-flowed the
+  numbering - and that is precisely why the older drift survived.
+
+### Fixed
+
+- **Nine drifted Locked citations in `analysis-template.md`.** It cited 13 for
+  paired goals (14), 12 for Gherkin (13), 14 for the SVG default (15), 16 for
+  exhaustive response variants (17), 21 for the concept layer (22), 28 for the
+  SwiftUI preview (27), 18 for variant drilling (19), and "17 + 18" for the
+  design reference (18 + 19). A tenth cited Locked 17 for analytics PII, which
+  no decision covers at all - the rule stays, the number goes, because a number
+  a reader cannot look up is worse than none. The new anchors hold all of them.
+- **Locked 33 was enforced all along and documented as prose.**
+  `build-references.mjs --check` runs its coverage gate and
+  `smoke-build-references.sh` tests it, but neither named the decision in a
+  failure message and the attribution check reads messages. Both name it now.
+  The same trap caught the new Locked 21 check on its first run: the attribution
+  scan reads an error message up to the first semicolon, and the message had one
+  in the middle.
+- **write-state verifies the write after the rename, not only before it.** No
+  POSIX call renames a file conditionally on still holding a lock, so the
+  `stillOurs` check is a time-of-check and the rename is the time-of-use. The
+  writer now reads the file back and compares the `rev` on disk against the one
+  it just wrote; a different rev means another writer's rename landed on top,
+  and the writer exits 4 instead of 0. Reporting success while losing an update
+  is the one outcome that script exists to prevent, and a window it could not
+  see was the one place that could still happen. The clobber branch has no test:
+  staging it from a shell needs a hook inside the writer, and a test-only hook
+  in the file that guards state is the worse trade.
+
+- **Stale phase numbers in eleven files.** `/multi-agent:review` recorded its
+  standalone runs under phase id 4; a tracker example in `phase-3-review.md`
+  drew `Phase 3  Dev`; `rules/figma-pipeline.md` carried the whole eight-phase
+  access matrix, including two rows for phases that no longer exist. Historical
+  files (CHANGELOG, ROADMAP entries, ADRs, migration headers) were left alone -
+  they narrate what was true then - and so were the separate phase namespaces
+  that `analysis/SKILL.md` and the Figma component flow use.
+
+### Changed
+
+- Toolkit counts follow `multi-agent-toolkit-mcp` 3.13.0: **115 tools in 13
+  categories**, up from 99 in 10. The new families are a full-text context index
+  (FTS5 + bm25 over offloaded payloads), provider-backed research, and video key
+  frames. `docs/ecosystem.md` gained their rows, `docs/facts.json` regenerated,
+  and the MCP-server context-cost note in `doctor` and the prefs schema now
+  names the real number.
+- `signal-community` records the second search path. The community-signal skill
+  carried a parity exemption because web search is not guaranteed on every host;
+  `research_search` runs over the MCP channel every host already speaks, so the
+  exemption now names the host's own search specifically and points at the
+  toolkit path as the preferred one.
+
+---
+
 ## [19.0.0] - 2026-09-18
 
 Major, because phase numbers are the contract and they moved. Eight phases

package/README.md

@@ -17,7 +17,7 @@ Runs natively on Claude Code, Copilot CLI and Codex CLI. macOS only. Zero runtim
 ### Prerequisites
 
 - **Node.js >= 20.11** - required; the pipeline's own tooling runs on it.
-- **`jq`** - required for nine paths, optional for the rest. 82 shell files call it. The nine that publish or decide - the autopilot queue, Jira comments, PR reviews, issue updates, the plan file, both Figma fetchers, log search and Jira auth - now refuse with exit 3 rather than run, because a missing `jq` renders as empty DATA and the work carries on with it. Everywhere else it still degrades. The install prints a note when it is missing.
+- **`jq`** - required for nine paths, optional for the rest. 84 shell files call it. The nine that publish or decide - the autopilot queue, Jira comments, PR reviews, issue updates, the plan file, both Figma fetchers, log search and Jira auth - now refuse with exit 3 rather than run, because a missing `jq` renders as empty DATA and the work carries on with it. Everywhere else it still degrades. The install prints a note when it is missing.
 - **`gh`** - for GitHub issue and PR work. Its built-in `--jq` is independent of the `jq` binary.
 
 ## Quick Start
@@ -134,7 +134,7 @@ Reasoning, mapping and rejected alternatives:
 
 Three smaller things in 17.6.0, each closing a gap where the pipeline assumed instead of looking:
 
-- **Phase 3 stopped typing `npm`.** A repo on pnpm, yarn or bun used to fail in the development phase, with a worktree and a branch already created. The manager is resolved from the repo now - an env override, then `package.json#packageManager`, then the lock file, then npm reported as a default rather than as evidence. iOS and Android are untouched.
+- **Phase 2 stopped typing `npm`.** A repo on pnpm, yarn or bun used to fail in the development phase, with a worktree and a branch already created. The manager is resolved from the repo now - an env override, then `package.json#packageManager`, then the lock file, then npm reported as a default rather than as evidence. iOS and Android are untouched.
 - **A compaction no longer eats what a phase learned.** The capture hook ran at session end; an auto-compaction summarizes a long review or development phase while it is still running, and anything not yet written was gone before session end ever fired. The hooks template now flushes at `PreCompact` too.
 - **`doctor` counts your MCP servers.** Every registered server sends its tool list on every turn and they are added one at a time, so nobody ever sees the total. It reports the count and nothing else: no warning, no blocking, no disabling.
 
@@ -246,7 +246,7 @@ The widget follows the answer rather than predicting it: Phase 0 is the only til
 | `/multi-agent:test-accessibility`        | VoiceOver labels, sub-44pt tap targets, contrast, traits                  |
 | `/multi-agent:test-dynamic-type`         | Re-walk every screen at XL through accessibility-XL, report truncation    |
 | `/multi-agent:test-screenshots [locale]` | App Store screenshot set in a locale (defaults to `tr`)                   |
-| `/multi-agent:manual-test`               | Phase 5 standalone: check out the task branch and prepare it for Xcode    |
+| `/multi-agent:manual-test`               | Phase 3 standalone: check out the task branch and prepare it for Xcode    |
 
 ### Design, build and store
 

package/docs/ecosystem.md

@@ -7,7 +7,7 @@ separately, wired together at install time and at run time:
 |---|---|---|
 | **`multi-agent-pipeline`** (this repo) | Orchestration: the 6-phase flow, the 60 slash commands, quality gates, review/triage, cross-CLI parity | npm package (`@mmerterden/multi-agent-pipeline`), installs itself onto Claude Code / Copilot CLI / Codex CLI |
 | **`multi-agent-plugins`** | Stack knowledge: per-platform component/lifecycle skills (iOS, Android, Frontend, Backend) + shared knowledge | Claude Code marketplace, 5 independently-versioned plugins |
-| **`multi-agent-toolkit-mcp`** | The pipeline's hands on devices and browsers: 99 MCP tools across 10 categories (simulator/emulator control, memory, crash diagnostics, accessibility audit, store compliance, web automation, Figma-vs-mock design audit, code intelligence, wallet passes, an agent-DSL batch runner) | npm package, registered as a standard stdio MCP server on every host |
+| **`multi-agent-toolkit-mcp`** | The pipeline's hands on devices and browsers: 115 MCP tools across 13 categories (simulator/emulator control, memory, crash diagnostics, accessibility audit, store compliance, web automation, Figma-vs-mock design audit, code intelligence, wallet passes, an agent-DSL batch runner, a full-text context index, provider-backed research, video key frames) | npm package, registered as a standard stdio MCP server on every host |
 
 None of the three depends on the others at the code level. They compose through two
 narrow contracts: the **Skill tool** (pipeline → plugin, at Phase 2) and the **MCP
@@ -211,12 +211,17 @@ that step's own contract).
 
 ### multi-agent-toolkit-mcp's tools, by category
 
-99 tools in 10 categories, counted from the server's own `tools/list` response at
-toolkit 3.12.0 rather than from a README. The table stood at 87 across 8
+115 tools in 13 categories, counted from the server's own `tools/list` response
+at toolkit 3.13.0 rather than from a README. The table stood at 87 across 8
 categories for several releases because Code Intelligence and Wallet Passes
 shipped without anyone adding their rows, and nothing here was checked against
 the server - which is why the count now names its source.
 
+The families can be narrowed per host: `MCP_TOOLKIT_CAPS` serves only the named
+ones, and unset serves all 115. A frontend repo that sets `web,code,context`
+sees 29 tools instead of 115, which matters because every registered server's
+tool list is charged against the context window on every turn.
+
 | Category | Tools | Primary pipeline consumers |
 |---|---|---|
 | Device Control | 59 | `/multi-agent:test`, `test-dark-mode`, `test-accessibility`, `test-dynamic-type`, `test-screenshots`, `manual-test`, `design-check` |
@@ -224,11 +229,14 @@ the server - which is why the count now names its source.
 | Crash Diagnostics | 2 (`ios_list_crashes`, `android_list_crashes`) | `/multi-agent:test` full scenario (end-of-run crash sweep), outside-the-pipeline sessions |
 | Accessibility Audit | 3 (`ios_accessibility_audit`, `android_accessibility_audit`, `ios_accessibility_audit_deep`) | `/multi-agent:test` accessibility scenario, `test-accessibility` |
 | Store Compliance | 5 | `store-ready`, `testflight-validation`, `apple-archive-compliance` skill, Phase 3 Security Auditor |
-| Web Automation | 8 | frontend-stack UI testing (via `test`) |
+| Web Automation | 18 | frontend-stack UI testing (via `test`); snapshot `ref` targeting, console/network capture, readable extraction and a bounded same-origin crawl |
 | Design Audit | 6 | `design-check` (mock-mode vs Figma conformance) |
 | Autonomous Agent DSL | 2 (`agent_run_steps`, `agent_query_output`) | any skill that needs a scripted multi-step device flow in one round trip |
 | Code Intelligence | 8 (`code_definition`, `code_references`, `code_hover`, `code_diagnostics`, `code_document_symbols`, `code_workspace_symbols`, `code_index_status`, `code_server_reset`) | `/multi-agent:refactor`, Phase 3 reviewers needing a real symbol graph rather than grep |
 | Wallet Passes | 4 (`pass_build`, `pass_validate`, `pass_inspect`, `pass_certificates`) | pass-kit work outside the pipeline; no pipeline phase consumes them |
+| Context Index | 3 (`context_index`, `context_search`, `context_get`) | any skill holding an offloaded payload too large to read; ranks passages instead of grepping lines |
+| Research | 2 (`research_search`, `research_ask`) | closes the parity exemption that had web search outside the MCP channel |
+| Media | 1 (`media_frames`) | Phase 2 visual evidence: turns a screen recording into frames a reviewer and a model can both look at |
 
 ---
 

package/docs/facts.json

@@ -1,7 +1,7 @@
 {
   "$comment": "Generated by pipeline/scripts/gen-facts.mjs. Do not hand-edit: the site reads this, and a number edited here instead of at its source is the drift this file removes.",
-  "generatedAt": "2026-09-17",
-  "version": "19.0.0",
+  "generatedAt": "2026-09-20",
+  "version": "19.1.0",
   "phaseSchema": 2,
   "phases": [
     {
@@ -40,6 +40,22 @@
   },
   "commandCount": 60,
   "skillCount": 153,
-  "toolCount": 99,
-  "toolkitVersion": "3.12.0"
+  "skillCountAll": 216,
+  "agentCount": 9,
+  "hosts": ["Claude Code", "Codex CLI", "Copilot CLI"],
+  "toolCount": 115,
+  "toolCategories": {
+    "ios": 40,
+    "android": 31,
+    "web": 18,
+    "code": 8,
+    "design": 6,
+    "pass": 4,
+    "context": 3,
+    "agent": 2,
+    "research": 2,
+    "media": 1
+  },
+  "pluginSkillCount": 238,
+  "toolkitVersion": "3.13.0"
 }

package/docs/features.md

@@ -39,7 +39,7 @@ The install is not only useful while `/multi-agent` is running. `rules/outside-t
 
 - **Onboarded service credentials.** Resolve the logical name through `credential-store.sh` and read the issue, page or log. Writes route through the pipeline commands, which carry the rules that make them safe - issues are never auto-closed, PR bodies use `Ref:`, outward prose goes through the humanizer.
 - **The stack skills enabled for this repo.** Each toolkit's own `index` skill routes. The pipeline reads the effective `enabledPlugins` rather than keeping a stack table, so a seventh toolkit needs no code change.
-- **The `multi-agent-toolkit` MCP.** 99 tools for a running app.
+- **The `multi-agent-toolkit` MCP.** 115 tools for a running app.
 
 Uninstall preserves the whole layer - tokens, the reader that opens them, the mapping that names them, the MCP registration. It is 1.5 kB of always-loaded text; the detail lives in a ref that loads on demand, and a gate keeps both under a ceiling because every byte there is paid by every session.
 

package/docs/recovery-guide.md

@@ -9,7 +9,7 @@ the answer across `modes.md`, `operations.md`, and the individual phase specs.
 | Symptom                                                 | Jump to                                         |
 | ------------------------------------------------------- | ----------------------------------------------- |
 | Pipeline paused/halted mid-run                          | [Resume a paused task](#resume-a-paused-task)   |
-| Phase 3 build failed > 3 times                          | [Build retry exhausted](#build-retry-exhausted) |
+| Phase 2 build failed > 3 times                          | [Build retry exhausted](#build-retry-exhausted) |
 | Phase 3 triage returned exit 1 (invalid JSON) twice     | [Triage fallback](#triage-fallback)             |
 | Phase 3 triage returned exit 2 (over-rejection)         | [Over-rejection](#over-rejection-guard)         |
 | Worktree already exists / dirty                         | [Worktree collisions](#worktree-collisions)     |
@@ -47,7 +47,7 @@ so `resume` continues in autopilot.
 
 ## Build Retry Exhausted
 
-Phase 3 retries the build up to **3 times** on each TDD cycle's green step. On
+Phase 2 retries the build up to **3 times** on each TDD cycle's green step. On
 the 4th consecutive failure the pipeline **pauses** (even in autopilot) and
 surfaces the error. Autopilot does not loop infinitely - this is intentional.
 
@@ -57,7 +57,7 @@ Path forward:
 2. If the failure is environmental (missing Xcode sim, wrong JDK), fix the
    environment and `resume` - the 3-retry counter resets.
 3. If the failure is logic (compile error in the generated code), edit the
-   offending file in the worktree, then `resume` - Phase 3 re-runs build.
+   offending file in the worktree, then `resume` - Phase 2 re-runs build.
 4. If the task itself is wrong-shaped (Phase 1 plan is infeasible), `kill` and
    restart with a better-scoped prompt.
 
@@ -69,15 +69,15 @@ Path forward:
 
 | Exit | Meaning                  | Pipeline action                                |
 | ---- | ------------------------ | ---------------------------------------------- |
-| 0    | Valid & clean            | Proceed to Phase 5                             |
+| 0    | Valid & clean            | Proceed to Phase 4                             |
 | 1    | Invalid JSON / schema    | Retry triage ONCE. On second exit-1, fallback: treat ALL raw findings as accepted. |
 | 2    | Over-rejection (>80%)    | Pause for human confirmation                   |
 | 3    | Contradiction corrected  | Use `result.corrected`, continue               |
 
 If exit 1 fires twice (fallback path):
 
-1. The pipeline logs `Phase 4: triage failed twice - fallback, all findings accepted as blocking`.
-2. All raw findings loop back into Phase 3 as if they were all real blockers.
+1. The pipeline logs `Phase 3: triage failed twice - fallback, all findings accepted as blocking`.
+2. All raw findings loop back into Phase 2 as if they were all real blockers.
 3. This is intentionally conservative - we'd rather over-fix than skip
    something real. You can manually mark noise in the Phase 4 PR description.
 
@@ -91,8 +91,8 @@ asks a human to look at the raw findings vs triage output.
 
 - Open `agent-log.md`, find the triage invocation.
 - Compare raw findings (logged) with rejected items.
-- If the rejections are correct, run `resume` with `--accept-overrejection` (Phase 4 re-enters validation with the guard soft-failed).
-- If not, run `resume` - Phase 4 re-runs triage with the guard still active.
+- If the rejections are correct, run `resume` with `--accept-overrejection` (Phase 3 re-enters validation with the guard soft-failed).
+- If not, run `resume` - Phase 3 re-runs triage with the guard still active.
 
 Autopilot behavior: logs the warning, accepts the triage output, continues -
 the autopilot contract values throughput over human-in-loop.