immune-brain 3.6.8 → 3.6.9

package/README.md

@@ -25,7 +25,7 @@ Pi and Claude Code are the supported hosts. Undeclared adapters remain unsupport
 - [Installation](#installation)
 - [Quick Start](#quick-start)
 - [How to Use](#how-to-use)
-- [The 6 Skills](#the-6-skills)
+- [The 7 Skills](#the-7-skills)
 - [Lifecycle](#lifecycle)
 - [Unattended Batch Runs](#unattended-batch-runs)
 - [Configuration](#configuration)
@@ -119,6 +119,7 @@ Immune-Brain provides two clean modes: **Host-native** for daily coding, and **M
 | PR has review comments or failing CI | `/imm-pr-fix` on that PR | → Standalone repair: minimal scoped fix in place, no managed task created |
 | Project docs out of date | `/imm-doc-prune` | → Read-only audit; deletes only user-approved stale docs from manifest |
 | Agent instructions bloated | `/imm-agent-doc-maintain` | → Minimizes tracked `AGENTS.md` / `CLAUDE.md` to essential non-discoverable rules |
+| Which model's edits keep coming back for review | `/imm-review-retro` | → Ranks models by review load and reports project usage from session logs |
 
 > **Core Principle: Skill-Explicit Entry**
 > - **Ordinary input stays host-native**: Natural language queries never automatically start planning or task enrollment. You choose when to turn on engineering rigor.
@@ -126,7 +127,7 @@ Immune-Brain provides two clean modes: **Host-native** for daily coding, and **M
 
 ---
 
-## The 6 Skills
+## The 7 Skills
 
 | Skill | Type | When to use | What it does |
 |---|---|---|---|
@@ -136,10 +137,11 @@ Immune-Brain provides two clean modes: **Host-native** for daily coding, and **M
 | `imm-pr-fix` | Standalone | CI failed / review comments on a PR | Repairs one PR in place, no managed authority |
 | `imm-doc-prune` | Standalone | Stale current docs | Deletes only the hash-approved manifest entries |
 | `imm-agent-doc-maintain` | Standalone | Bloated agent instructions | Minimizes tracked AGENTS/CLAUDE/GEMINI.md to necessary context |
+| `imm-review-retro` | Standalone | Compare models by review load | Ranks authors of reviewed code and reports project usage |
 
 Internal roles (Executor, QA, Review, Compounder) are dispatched by `imm-loop` — you never invoke them directly.
 
-All 6 skills are invoked explicitly. For new features, start with `imm-brainstorm` (if requirements are uncertain) or `imm-planner` (if requirements are clear), then proceed to `imm-loop` once enrolled.
+All 7 skills are invoked explicitly. For new features, start with `imm-brainstorm` (if requirements are uncertain) or `imm-planner` (if requirements are clear), then proceed to `imm-loop` once enrolled.
 
 ### Managed Path entries (brainstorm → planner → loop)
 
@@ -187,6 +189,11 @@ The three repair/maintenance skills are host-native: they never create a managed
 - **Trigger:** explicit request to minimize tracked `AGENTS.md` / `CLAUDE.md` / `GEMINI.md`.
 - **What it does:** keeps only the necessary non-discoverable rules in agent instruction files, under the same read-only-audit + hash-bound-manifest-approval model as `imm-doc-prune`.
 
+#### `imm-review-retro` — review load and project usage
+
+- **Trigger:** explicit request for a cross-model review retro or project usage look-back.
+- **What it does:** ranks models by how much review their own edits triggered, plus sessions/turns/edits/tool mix, from pi session logs. Read-only. Not a diff review.
+
 ---
 
 ## Lifecycle
@@ -265,7 +272,7 @@ See [`docs/reference/immune-brain-config.md`](docs/reference/immune-brain-config
 package.json                          # Pi package manifest (skills + extensions)
 plugins/immune-brain/
 ├── .pi-extension/                    # Pi TUI + Kernel authority extension
-├── skills/                           # 6 public Skills (trigger shims)
+├── skills/                           # 7 public Skills (trigger shims)
 ├── dist/                             # Built skill contracts & references
 ├── runtime/                          # Bun + TypeScript runtime & Kernel
 └── bin/                              # CLI wrappers (→ runtime/v4_runtime.ts)

package/README.zh-CN.md

@@ -25,7 +25,7 @@ Pi 与 Claude Code 是支持的宿主。未声明的适配器仍不受支持。C
 - [安装](#安装)
 - [快速开始](#快速开始)
 - [如何使用](#如何使用)
-- [6 个 Skills](#6-个-skills)
+- [7 个 Skills](#7-个-skills)
 - [生命周期](#生命周期)
 - [无人值守批次运行](#无人值守批次运行)
 - [配置](#配置)
@@ -119,6 +119,7 @@ Immune-Brain 提供两种清晰的工作模式：日常轻量编码走 **Host-na
 | PR 被评论 / CI 挂了 | 对该 PR 使用 `/imm-pr-fix` | → 独立修复：在当前 PR 内针对性修复，不创建新 managed 任务 |
 | 文档过时需要清理 | `/imm-doc-prune` | → 只读审计过时文档，仅删除经哈希审批的条目 |
 | Agent 指令文件膨胀 | `/imm-agent-doc-maintain` | → 将 tracked `AGENTS.md` / `CLAUDE.md` 压到最小必要上下文 |
+| 想知道哪个模型的改动总被审查 | `/imm-review-retro` | → 按模型排名审查负载，并从 session logs 汇报项目使用量 |
 
 > **核心原则：Skill 显式调用**
 > - **普通输入保持 Host-native**：自然语言提问绝不自动绑架流程或发起 Enrollment。你完全自主决定何时开启严格工程保障。
@@ -126,7 +127,7 @@ Immune-Brain 提供两种清晰的工作模式：日常轻量编码走 **Host-na
 
 ---
 
-## 6 个 Skills
+## 7 个 Skills
 
 | Skill | 类型 | 何时使用 | 职责 |
 |---|---|---|---|
@@ -136,10 +137,11 @@ Immune-Brain 提供两种清晰的工作模式：日常轻量编码走 **Host-na
 | `imm-pr-fix` | 独立 | PR 需修复 | 原地修复单个 PR，不触及 managed authority |
 | `imm-doc-prune` | 独立 | 清理过时文档 | 仅删除哈希绑定的 manifest 条目 |
 | `imm-agent-doc-maintain` | 独立 | Agent instruction 膨胀 | 将 tracked AGENTS/CLAUDE/GEMINI.md 压到最小必要上下文 |
+| `imm-review-retro` | 独立 | 比较模型的审查负载 | 排名被审查代码的作者并汇报项目使用量 |
 
 Executor、QA、Review、Compounder 等为 `imm-loop` 内部调度的角色，无需手动调用。
 
-所有 6 个 Skill 均显式调用。新需求开发时：若需求含糊先调 `imm-brainstorm`，目标清晰直接调 `imm-planner`，完成确认后调 `imm-loop` 推进闭环。
+所有 7 个 Skill 均显式调用。新需求开发时：若需求含糊先调 `imm-brainstorm`，目标清晰直接调 `imm-planner`，完成确认后调 `imm-loop` 推进闭环。
 
 ### Managed Path 入口（brainstorm → planner → loop）
 
@@ -187,6 +189,11 @@ Executor、QA、Review、Compounder 等为 `imm-loop` 内部调度的角色，
 - **触发方式：** 显式要求精简版本控制下的 `AGENTS.md` / `CLAUDE.md` / `GEMINI.md`。
 - **职责：** 遵循与 `imm-doc-prune` 相同的「只读审计 + 哈希清单审批」模式，仅保留无法直接推导的必要规则。
 
+#### `imm-review-retro` — 审查负载与项目使用回顾
+
+- **触发方式：** 显式要求跨模型审查复盘或项目使用量回顾。
+- **职责：** 从 pi session logs 按模型排名被审查代码的作者，并汇报 sessions/turns/编辑量/工具分布。只读；不审查 diff。
+
 ---
 
 ## 生命周期

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "immune-brain",
-  "version": "3.6.8",
+  "version": "3.6.9",
   "description": "Immune-Brain agent skill system",
   "publishConfig": {
     "access": "public",

package/plugins/immune-brain/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "immune-brain",
-  "version": "3.6.8",
+  "version": "3.6.9",
   "description": "Immune-Brain Claude Code Host: native Enrollment, QA, Review, and Kernel settlement.",
   "author": {
     "name": "Immune-Brain Team"

package/plugins/immune-brain/dist/claude/mcp-server.mjs

@@ -42,7 +42,7 @@ function probeHost(env = process.env, platform = process.platform, hostVersion)
 }
 
 // plugins/immune-brain/runtime/plugin_version.ts
-var PLUGIN_VERSION = "3.6.8";
+var PLUGIN_VERSION = "3.6.9";
 
 // plugins/immune-brain/runtime/claude/interaction.ts
 import { createHash, randomUUID } from "node:crypto";
@@ -7126,7 +7126,8 @@ function parseIssues(raw) {
       title: item.title,
       body: typeof item.body === "string" ? item.body : "",
       state: item.state,
-      state_reason: typeof item.state_reason === "string" ? item.state_reason.toLowerCase() : null
+      state_reason: typeof item.state_reason === "string" ? item.state_reason.toLowerCase() : null,
+      labels: Array.isArray(item.labels) ? item.labels.map((label) => typeof label === "string" ? label : label?.name).filter((name) => typeof name === "string") : []
     };
   });
 }

package/plugins/immune-brain/dist/imm-planner.md

@@ -172,8 +172,13 @@ with `valid: true` and `enrollment_ready: true`. Resolve `../bin/imm-tracker` fr
 Initiative slug and goal, Parent projection, and every Child's `slice_id`,
 canonical TaskIntent path, bounded public `acceptance` summaries, and public
 projection. The Parent projection requires
-`problem`, `result`, and `design`, and may include `decisions`,
-`testing_strategy`, and `out_of_scope`. `design` records Initiative-level
+`short_name`, `title`, `problem`, `result`, and `design`, and may include
+`source_issue`, `decisions`,
+`testing_strategy`, and `out_of_scope`. `short_name` (1-32 characters) is the
+stable short Initiative name used in every Issue title; `title` (1-60
+characters) is the short Initiative display title; `source_issue` is the
+originating feature Issue number, rendered as a Provenance link. `design`
+records Initiative-level
 invariants, Slice boundaries and ordering, shared interfaces or state flow, and
 material compatibility decisions. Every Parent Slice must correspond to one
 published Child; future checklist-only Slices are not allowed in the batch.
@@ -181,15 +186,24 @@ published Child; future checklist-only Slices are not allowed in the batch.
 Each Child must provide public `acceptance` entries with `id` and a 1-500
 character `summary`. Their IDs must match every canonical TaskIntent acceptance
 ID exactly once. Canonical assertion prose is authority evidence and must never
-be copied into public GitHub projection. Each Child projection may contain
+be copied into public GitHub projection. Each Child projection requires
+`title` (1-60 characters), the short Slice display title, and may contain
 `result`, `current_behavior`,
 `desired_behavior`, `key_interfaces`, `verification`, `blocked_by` Task IDs,
-`out_of_scope`, and `agent_handoff`. The tracker rereads every canonical
+`out_of_scope`, and `agent_handoff`. The tracker composes Issue titles from
+these display names only — the Parent as `[<short_name>] <title>` and each
+Child as `[<short_name>] S<n> <title>` with `n` the declared Slice position —
+and fails the whole batch closed before any remote write when a display name
+is missing or the composed title exceeds 80 characters; it never falls back to
+goal prose and never truncates a title. The tracker rereads every canonical
 TaskIntent for identity, risk, and acceptance IDs; projection fields and public
 summaries never widen TaskIntent scope or authority. It validates the complete dependency graph before
 remote writes, creates the Parent once, creates all Children, attaches every
 Child as a native Sub-issue, creates native `blocked_by` relations, and rereads
-the complete topology. The Child Agent Brief includes a direct Parent Issue link.
+the complete topology. Every Child carries `ready-for-agent`, blocked Children
+additionally carry `blocked`, and the Parent carries neither; the tracker never
+creates labels, so a repository missing a required label fails the batch closed
+before any remote write. The Child Agent Brief includes a direct Parent Issue link.
 Internal role prompts, tool policies, review gates, model reservations, and
 prompt digests never belong in this external handoff. If
 `docs/initiatives/<slug>.md` exists, publication fails with a carrier conflict;

package/plugins/immune-brain/dist/imm-review-retro.md

@@ -0,0 +1,123 @@
+---
+name: imm-review-retro
+description: Use when the user explicitly requests Immune-Brain ranking of models by cross-model review load or a project usage retro.
+---
+
+# Immune-Brain: Review Retro
+
+Rank models by how much code review their own edits triggered, and report
+basic project usage over a look-back window the user supplies in days. This
+is a standalone host-native analysis entry, not a Managed Path continuation
+and not an `imm-loop` internal-role dispatch. It reviews no diff — a diff
+review is `code-review`.
+
+## Boundary
+
+Allowed: read pi session JSONL under `~/.pi/agent/sessions` (or `--root`),
+run the bundled analyzer, and write a stdout report.
+
+Blocked: code, test, Spec, Plan, or `.imm/` edits; session-log writes;
+Kernel, TaskIntent, or TaskRecord mutation; Compounder or scheduled runs;
+`.imm/audit/` lifecycle statistics.
+
+An already active Managed task remains owned by `imm-loop`. This Skill does
+not create or resume Managed authority.
+
+## Invocation
+
+Requires explicit invocation: `imm-review-retro` or `/imm-review-retro`.
+Ordinary questions such as "which model is worse" stay host-native and do
+not enter this Skill.
+
+The look-back window in days is required input. If the user named one, use
+it. If not, ask before running, because the ranking moves with the window.
+
+Default scan is the user's full session-log tree. Pass `--project <substr>`
+when the user wants one repo or worktree. Do not invent a project filter.
+
+No daemon, no cron, no CI, no automatic commit.
+
+## Counting rules
+
+These rules keep numbers comparable across runs. Read the analyzer header
+aloud in the report so the 口径 stays visible.
+
+- `review` = an `Agent` tool call with `subagent_type` equal to `Review`.
+- Attribution = the model behind the most recent `edit`, `write`, or
+  `multiedit` in that session. If none, the row is `no-edit (review-only)`.
+- `uniq` counts distinct (session, description+prompt prefix) pairs. A wide
+  gap versus `reviews` is the same review re-run on the same code.
+- `rev/100ed` is `100 * reviews / devEdits`. Rank on both absolute `reviews`
+  and this intensity. A model can lead one axis and sit mid-pack on the
+  other.
+- `avgSc` / `pass%` parse `[SCORE: …]` and `[VERDICT: …]` tags from the
+  matching Review `toolResult`. Untagged reviews show `-`.
+- `registr` counts `imm_kernel_canary` `submit_review`. It is the
+  registration of the same review and is never added into `reviews`.
+- `rounds/task` is registrations per distinct `(cwd, task_id)`. High values
+  can be canary/QA harness re-registration, not human-visible rework.
+- Findings are `record_finding` calls, deduped per session. Summaries that
+  match `recorded cleanly`, `receipt recorded`, `round recorded`, or
+  `no finding(s)` are `bookkeep` / `noisy`, excluded from `block`/`advis`.
+
+Usage counters on the same pass: sessions with activity, assistant turns,
+edit counts, a tool-call name histogram, and the project × author table.
+
+## CLI
+
+Run the bundled analyzer. Prefer `bun`; `node` (≥23.6, type stripping) is
+an allowed equivalent. The script is erasable TypeScript with `node:` APIs
+only.
+
+```
+bun "<path-to-skill>/scripts/review_retro.ts" <days> [--root <sessions-dir>] [--project <substr>] [--top N]
+```
+
+- `<days>` must be `> 0`.
+- `--root` defaults to `~/.pi/agent/sessions`.
+- `--project` keeps sessions whose `cwd` contains the substring.
+- `--top` is the project-table row cap (default 15).
+- Malformed JSONL lines are skipped. `days <= 0` is a hard error.
+
+Do not scan live `.imm/` directories. Tests use committed fixtures under
+`tests/fixtures/review-retro/`.
+
+## Report
+
+Write-up order:
+
+1. Window and 口径 in one line (copy the analyzer header).
+2. Ranked model table, including scores.
+3. Usage section: sessions, turns, edits, tool mix.
+4. Quality and score findings.
+5. Three to five bullets of what the table means (volume versus intensity,
+   quality versus rework, where it concentrated).
+6. Caveats last.
+
+Rank on both axes, never one. Name the axis you are ranking by, and call
+out models that flip order between `reviews` and `rev/100ed`.
+
+Separate one-pass from rework: compare `uniq` to `reviews`, and read
+`rounds/task` on the kernel path.
+
+Evaluate quality: high intensity plus high score is frequent review of
+mostly minor issues; low intensity plus low score is rare review of severe
+defects. Call out REJECT or highRisk ratings.
+
+Ground each model in its projects. Cite the two or three worktrees where
+that model's reviews concentrated.
+
+## Caveats
+
+Anything the script splits out as `bookkeep` stays visible next to the
+column it contaminates. Flag any finding count you cannot trace to a real
+defect.
+
+High `rounds/task` can be canary/QA harness re-registration, not
+human-visible rework.
+
+This Skill does not persist snapshots or compute week-over-week diffs.
+Re-run with a new window when the user wants a later period.
+
+The personal python prototype under `~/.pi/agent/skills/review-retro/` is
+not this Skill and is not modified by it.

package/plugins/immune-brain/dist/registry.yaml

@@ -56,3 +56,12 @@ skills:
     output_artifacts: [maintain_report]
     next_actions: []
     boundary: Minimize tracked agent-instruction context after explicit manifest approval; no Managed authority mutation, contract installation, or reference-document creation.
+  - name: imm-review-retro
+    path: skills/imm-review-retro/SKILL.md
+    role: execute
+    title: Review Retro
+    role_class: discovery
+    canonical: true
+    output_artifacts: [retro_report]
+    next_actions: []
+    boundary: Rank models by review load and report project usage from pi session logs. Read-only. No Managed Path mutation.

package/plugins/immune-brain/runtime/github_issue_tracker.ts

@@ -14,6 +14,15 @@ const GH_TIMEOUT_MS = 20_000;
 const MAX_SNAPSHOT_PAGES = 100;
 const GITHUB_ISSUE_BODY_LIMIT = 65_536;
 const MAX_TERMINAL_EVENT_ID = 500;
+const MAX_TITLE_LENGTH = 80;
+const MAX_DISPLAY_SHORT_NAME = 32;
+const MAX_DISPLAY_TITLE = 60;
+const READY_FOR_AGENT_LABEL = "ready-for-agent";
+const BLOCKED_LABEL = "blocked";
+const MANAGED_ISSUE_LABELS: readonly string[] = [READY_FOR_AGENT_LABEL, BLOCKED_LABEL];
+
+/** Single compressed Activity/Authority footer shared by Parent and Task Issues. */
+const ISSUE_FOOTER = "---\n\n_Outbound visibility only: GitHub state never authorizes or settles work — Kernel TaskIntent, TaskRecord, QA, Review, and Assurance remain the execution authority. An Open Issue only means the Task still needs attention; only a claimless terminal projection closes it (`done` → Completed, `stopped` → Not planned)._";
 
 type TrackerStatus =
 	| "created"
@@ -42,6 +51,9 @@ export interface InitiativeSlice {
 }
 
 export interface InitiativeProjection {
+	short_name?: string;
+	title?: string;
+	source_issue?: string;
 	problem?: string;
 	result?: string;
 	design?: string;
@@ -128,6 +140,9 @@ export interface GithubInitiativeObservation {
 }
 
 export interface TaskProjection {
+	short_name?: string;
+	title?: string;
+	slice_ordinal?: number;
 	result?: string;
 	current_behavior?: string;
 	desired_behavior?: string;
@@ -188,6 +203,7 @@ interface GithubIssue {
 	body: string;
 	state: "open" | "closed";
 	state_reason: string | null;
+	labels: string[];
 }
 
 interface RepositorySnapshot {
@@ -276,12 +292,120 @@ function titleText(value: string): string {
 	return redactSecrets(value).replace(/\s+/g, " ");
 }
 
-function issueTitle(owner: string, result: string): string {
-	const title = `[${owner}] ${titleText(result)}`;
-	if (title.length > 256) throw new Error("GitHub Issue title must not exceed 256 characters");
+/**
+ * A Planner-supplied display name for an Issue title. Display names are the
+ * only title source: the full goal/result prose stays in the body, and a
+ * missing or oversized display name fails closed instead of being truncated.
+ */
+function displayName(value: unknown, name: string, max: number): string {
+	if (typeof value !== "string" || !value.trim())
+		throw new Error(`${name} is required: publish a bounded display name instead of the full goal text`);
+	// Brackets are rejected on the raw value: redaction later introduces its own
+	// bracketed marker, which must never be mistaken for caller-supplied syntax.
+	if (/[[\]]/.test(value)) throw new Error(`${name} must not contain square brackets`);
+	return projectionText(value, name, max);
+}
+
+function issueTitle(value: string): string {
+	const title = titleText(value);
+	if (title.length > MAX_TITLE_LENGTH)
+		throw new Error(`GitHub Issue title must not exceed ${MAX_TITLE_LENGTH} characters: shorten the Planner display names`);
 	return title;
 }
 
+function initiativeDisplayNames(projection: InitiativeProjection | undefined): { shortName: string } {
+	const missing = (["short_name", "title"] as const).filter((field) => projection?.[field] === undefined);
+	if (missing.length)
+		throw new Error(`Initiative projection requires display names; missing ${missing.map((field) => `projection.${field}`).join(", ")}`);
+	return { shortName: displayName(projection?.short_name, "projection.short_name", MAX_DISPLAY_SHORT_NAME) };
+}
+
+function initiativeIssueTitle(initiativeId: string, projection: InitiativeProjection | undefined): string {
+	const { shortName } = initiativeDisplayNames(projection);
+	const title = displayName(projection?.title, "projection.title", MAX_DISPLAY_TITLE);
+	return issueTitle(`[${shortName}] ${title}`);
+}
+
+function taskDisplayNames(operation: Extract<TrackerOperation, { op: "upsert-task" }>): { shortName: string; title: string; ordinal: number } {
+	const projection = operation.projection;
+	if (projection?.short_name === undefined || projection.title === undefined || projection.slice_ordinal === undefined)
+		throw new Error(`Task ${operation.task_id} requires projection.short_name, projection.title, and projection.slice_ordinal display names`);
+	return {
+		shortName: displayName(projection.short_name, "projection.short_name", MAX_DISPLAY_SHORT_NAME),
+		title: displayName(projection.title, "projection.title", MAX_DISPLAY_TITLE),
+		ordinal: projection.slice_ordinal,
+	};
+}
+
+function taskIssueTitle(operation: Extract<TrackerOperation, { op: "upsert-task" }>, fallbackOrdinal?: number): string {
+	const { shortName, title, ordinal } = taskDisplayNames(operation);
+	return issueTitle(`[${shortName}] S${fallbackOrdinal ?? ordinal} ${title}`);
+}
+
+/**
+ * The Slice position in the Initiative: the Parent's Slices checklist is the
+ * declared order, so a Child keeps the same `S<n>` across amendment batches
+ * instead of renumbering to its position inside the current batch.
+ */
+function sliceOrdinalFromChecklist(parentBody: string, sliceId: string, fallback: number): number {
+	const declared = [...parentBody.matchAll(/^- \[[ xX]\] <!-- immune-brain:slice-id=([A-Za-z0-9._:-]+) -->/gm)].map((match) => match[1]);
+	const index = declared.indexOf(sliceId);
+	return index === -1 ? fallback : index + 1;
+}
+
+/** Labels a published Task Issue must carry; the Parent carries none of them. */
+function desiredTaskLabels(operation: Extract<TrackerOperation, { op: "upsert-task" }>): string[] {
+	return (operation.projection?.blocked_by ?? []).length
+		? [READY_FOR_AGENT_LABEL, BLOCKED_LABEL]
+		: [READY_FOR_AGENT_LABEL];
+}
+
+/** Converge managed labels from the observed set: add what is desired, remove only managed labels that are not. */
+function labelMutationArgs(observed: string[], desired: string[]): string[] {
+	const args: string[] = [];
+	for (const label of desired) if (!observed.includes(label)) args.push("--add-label", label);
+	for (const label of MANAGED_ISSUE_LABELS)
+		if (observed.includes(label) && !desired.includes(label)) args.push("--remove-label", label);
+	return args;
+}
+
+async function repositoryLabels(root: string, gh: GhTransport, repository: RepositoryInfo): Promise<string[] | GithubTrackerResult> {
+	const execution = await gh.run(
+		["label", "list", "--repo", repository.name_with_owner, "--json", "name", "--limit", "1000"],
+		{ cwd: root },
+	);
+	if (execution.exit_code !== 0 || execution.output_exceeded)
+		return ghFailure("upsert-task", execution, "cannot query repository labels");
+	try {
+		const parsed = JSON.parse(execution.stdout) as unknown;
+		if (!Array.isArray(parsed)) throw new Error("gh returned malformed label list");
+		return parsed
+			.map((item) => (item as { name?: unknown })?.name)
+			.filter((name): name is string => typeof name === "string");
+	} catch (error) {
+		return result("upsert-task", "permanent_failure", error instanceof Error ? error.message : String(error));
+	}
+}
+
+/**
+ * Labels are never created by the tracker: a repository missing a required
+ * label fails closed before any Issue mutation, naming the exact label.
+ */
+async function labelAvailabilityFailure(
+	root: string,
+	gh: GhTransport,
+	repository: RepositoryInfo,
+	required: string[],
+): Promise<GithubTrackerResult | null> {
+	if (!required.length) return null;
+	const labels = await repositoryLabels(root, gh, repository);
+	if (!Array.isArray(labels)) return labels;
+	const missing = required.filter((label) => !labels.includes(label));
+	return missing.length
+		? result("upsert-task", "permanent_failure", `repository labels missing: ${missing.join(", ")}; create them before publishing so Task Issues carry publication state`)
+		: null;
+}
+
 export function redactGithubDiagnostic(value: string): string {
 	return redactSecrets(value)
 		.replace(/\s+/g, " ")
@@ -414,6 +538,11 @@ function parseIssues(raw: string): GithubIssue[] {
 				body: typeof item.body === "string" ? item.body : "",
 				state: item.state,
 				state_reason: typeof item.state_reason === "string" ? item.state_reason.toLowerCase() : null,
+				labels: Array.isArray(item.labels)
+					? item.labels
+						.map((label) => (typeof label === "string" ? label : (label as { name?: unknown })?.name))
+						.filter((name): name is string => typeof name === "string")
+					: [],
 			};
 		});
 }
@@ -721,12 +850,15 @@ function createInitiativeBody(
 	historicalSlices: string[] = [],
 ): string {
 	const projection = operation.projection ?? {};
+	const provenance = projection.source_issue
+		? `## Provenance\n\n- Derived from #${projection.source_issue}: the originating feature Issue for this Initiative.\n\n`
+		: "";
 	return `${[
 		PROTOCOL_MARKER,
 		KIND_INITIATIVE_MARKER,
 		marker("repo-id", repository.id),
 		marker("initiative-id", operation.initiative_id),
-	].join("\n")}\n\n# ${titleText(projection.result ?? operation.goal)}\n\nOpt-in, non-authoritative Immune-Brain Initiative planning carrier. Kernel TaskIntent, TaskRecord, and Assurance remain the execution authority.\n\n## How to use this Issue\n\n- Edit planning prose and Slice ordering directly after creation.\n- Keep each Slice marker attached to exactly one stable Slice entry.\n- The tracker never rewrites or closes this Parent after creation; the tracker never changes or closes it automatically.\n\n## Problem\n\n${publicText(projection.problem ?? "The Initiative addresses the bounded delivery described below.", "projection.problem")}\n\n## Result\n\n${publicText(projection.result ?? operation.goal, "projection.result")}\n\n## Initiative design\n\n${publicText(projection.design ?? "Each Child preserves the shared Initiative decisions and boundaries recorded here.", "projection.design")}\n\n## Decisions\n\n${listText(projection.decisions, "- No additional Initiative decisions recorded.")}\n\n## Testing strategy\n\n${publicText(projection.testing_strategy ?? "Each Child closes from its focused acceptance verification.", "projection.testing_strategy")}\n\n## Out of scope\n\n${listText(projection.out_of_scope, "- Unrelated work outside this Initiative.")}\n\n## Slices\n\n${operation.slices.length + historicalSlices.length === 0 ? "No Slices recorded yet." : [...historicalSlices, ...operation.slices.map((slice) => `- [ ] ${marker("slice-id", slice.id)} **${slice.id}**: ${slice.result ?? slice.goal}${slice.blocked_by?.length ? ` (blocked by: ${slice.blocked_by.join(", ")})` : ""}`)].join("\n")}\n\n## Authority boundary\n\nThis Issue is outbound visibility only. GitHub state never starts, authorizes, reprioritizes, or settles work. Native Sub-issues identify published Tasks; their state is only an observation.\n`;
+	].join("\n")}\n\n${provenance}## How to use this Issue\n\n- Edit planning prose and Slice ordering directly after creation.\n- Keep each Slice marker attached to exactly one stable Slice entry.\n- The tracker never rewrites or closes this Parent after creation; the tracker never changes or closes it automatically.\n\n## Problem\n\n${publicText(projection.problem ?? "The Initiative addresses the bounded delivery described below.", "projection.problem")}\n\n## Result\n\n${publicText(projection.result ?? operation.goal, "projection.result")}\n\n## Initiative design\n\n${publicText(projection.design ?? "Each Child preserves the shared Initiative decisions and boundaries recorded here.", "projection.design")}\n\n## Decisions\n\n${listText(projection.decisions, "- No additional Initiative decisions recorded.")}\n\n## Testing strategy\n\n${publicText(projection.testing_strategy ?? "Each Child closes from its focused acceptance verification.", "projection.testing_strategy")}\n\n## Out of scope\n\n${listText(projection.out_of_scope, "- Unrelated work outside this Initiative.")}\n\n## Slices\n\n${operation.slices.length + historicalSlices.length === 0 ? "No Slices recorded yet." : [...historicalSlices, ...operation.slices.map((slice) => `- [ ] ${marker("slice-id", slice.id)} **${slice.id}**: ${slice.result ?? slice.goal}${slice.blocked_by?.length ? ` (blocked by: ${slice.blocked_by.join(", ")})` : ""}`)].join("\n")}\n\n${ISSUE_FOOTER}\n`;
 }
 
 async function createInitiative(
@@ -741,7 +873,7 @@ async function createInitiative(
 	const body = createInitiativeBody(source.repository, operation);
 	const oversized = bodyLimitFailure(operation.op, body);
 	if (oversized) return oversized;
-	const title = issueTitle(operation.initiative_id, operation.projection?.result ?? operation.goal);
+	const title = initiativeIssueTitle(operation.initiative_id, operation.projection);
 	if (found.kind === "missing") {
 		const mutation = await gh.run([
 			"issue", "create", "--repo", source.repository.name_with_owner,
@@ -761,8 +893,28 @@ async function createInitiative(
 			? result(operation.op, "created", "Initiative Issue created as the single GitHub source", confirmed.issue)
 			: result(operation.op, "retryable_failure", "Initiative Issue did not converge to the requested initial title and body", confirmed.issue);
 	}
-	if (found.issue.body === body && found.issue.title === title)
+	if (found.issue.body === body && found.issue.title === title) {
+		// A repeated complete batch is a convergence pass: managed labels the Parent
+		// must not carry are repaired here, without rewriting its content.
+		const labelArgs = labelMutationArgs(found.issue.labels, []);
+		if (labelArgs.length) {
+			const edited = await gh.run([
+				"issue", "edit", String(found.issue.number), "--repo", source.repository.name_with_owner,
+				...labelArgs,
+			], { cwd: root });
+			if (edited.exit_code !== 0 || edited.output_exceeded)
+				return ghFailure(operation.op, edited, "Initiative Parent label convergence failed");
+			const refreshed = await snapshot(root, gh, operation.op);
+			if ("contract" in refreshed) return refreshed;
+			const confirmed = lookup(refreshed.issues);
+			if (confirmed.kind !== "found")
+				return result(operation.op, "ambiguous_remote_state", "Initiative Parent became ambiguous after label convergence", found.issue);
+			return confirmed.issue.title === title && confirmed.issue.body === body && !labelMutationArgs(confirmed.issue.labels, []).length
+				? result(operation.op, "updated", "Initiative Issue managed labels converged", confirmed.issue)
+				: result(operation.op, "retryable_failure", "Initiative Issue did not converge to an unlabeled Parent", confirmed.issue);
+		}
 		return result(operation.op, "already_current", "Initiative Issue already carries the requested initial source", found.issue);
+	}
 	return result(
 		operation.op,
 		"permanent_failure",
@@ -785,7 +937,7 @@ function childBody(
 		marker("initiative-id", operation.initiative_id),
 		marker("slice-id", operation.slice_id),
 		marker("task-id", operation.task_id),
-	].join("\n")}\n\n# ${titleText(projection.result ?? operation.goal)}\n\nOpt-in, non-authoritative Immune-Brain Task Issue. Kernel TaskIntent, TaskRecord, and Assurance remain the execution authority.\n\n## Parent\n\n| Initiative | \`${operation.initiative_id}\` |\n| Parent Issue | [#${parent.number}](${parent.url}) |\n| Slice | \`${operation.slice_id}\` |\n| Risk | \`${operation.risk}\` |\n\n## What to build\n\n${publicText(projection.result ?? operation.goal, "projection.result")}\n\n## Current behavior\n\n${publicText(projection.current_behavior ?? "The current behavior is defined by the repository's existing contract.", "projection.current_behavior")}\n\n## Desired behavior\n\n${publicText(projection.desired_behavior ?? operation.goal, "projection.desired_behavior")}\n\n## Key interfaces\n\n${listText(projection.key_interfaces, "- Canonical TaskIntent acceptance and Kernel lifecycle remain authoritative.")}\n\n## Acceptance criteria\n\n${acceptance}\n\n## Verification\n\n${publicText(projection.verification ?? "Run the focused acceptance verification declared by the TaskIntent.", "projection.verification")}\n\n## Blocked by\n\n${projection.blocked_by?.length ? projection.blocked_by.map((id) => `- \`${identifier(id, "blocked_by task_id")}\``).join("\n") : "None"}\n\n## Out of scope\n\n${listText(projection.out_of_scope, "- Scope not declared by the validated TaskIntent.")}\n\n## Agent handoff\n\n${publicText(projection.agent_handoff ?? "Implement only the bounded TaskIntent result and run the focused checks. Do not widen scope or treat GitHub as authorization.", "projection.agent_handoff")}\n\n## Lifecycle\n\n- **Open** means this Task still needs attention; it does not mean the Task is authorized or executing.\n- Only a fresh claimless terminal projection can close this Issue: \`done\` becomes **Completed**, and \`stopped\` becomes **Not planned**.\n\n## Authority boundary\n\nThis Issue is outbound visibility only. GitHub state never changes TaskIntent, TaskRecord, QA, Review, authorization, or Kernel settlement. Internal role prompts, tool policies, review gates, model reservations, and prompt digests are not part of this external handoff.\n`;
+	].join("\n")}\n\n## Parent\n\n| Initiative | \`${operation.initiative_id}\` |\n| Parent Issue | [#${parent.number}](${parent.url}) |\n| Slice | \`${operation.slice_id}\` |\n| Risk | \`${operation.risk}\` |\n\n## Current behavior\n\n${publicText(projection.current_behavior ?? "The current behavior is defined by the repository's existing contract.", "projection.current_behavior")}\n\n## Desired behavior\n\n${publicText(projection.desired_behavior ?? projection.result ?? operation.goal, "projection.desired_behavior")}\n\n## Key interfaces\n\n${listText(projection.key_interfaces, "- Canonical TaskIntent acceptance and Kernel lifecycle remain authoritative.")}\n\n## Acceptance criteria\n\n${acceptance}\n\n## Verification\n\n${publicText(projection.verification ?? "Run the focused acceptance verification declared by the TaskIntent.", "projection.verification")}\n\n## Blocked by\n\n${projection.blocked_by?.length ? projection.blocked_by.map((id) => `- \`${identifier(id, "blocked_by task_id")}\``).join("\n") : "None"}\n\n## Out of scope\n\n${listText(projection.out_of_scope, "- Scope not declared by the validated TaskIntent.")}\n\n## Agent handoff\n\n${publicText(projection.agent_handoff ?? "Implement only the bounded TaskIntent result and run the focused checks. Do not widen scope or treat GitHub as authorization.", "projection.agent_handoff")}\n\n${ISSUE_FOOTER}\n`;
 }
 
 /** Derived approved-final content and baseline-derived historical evidence for an amendment. */
@@ -846,28 +998,30 @@ function approvedAmendmentContent(
 	);
 	const historicalSlices = baselineHistoricalSlices(amendment.parent.body, amendedSliceIds, boundSliceIds);
 	if (typeof historicalSlices === "string") return historicalSlices;
-	const pendingContent = new Map<string, { title: string; body: string }>();
 	const sourceStub = {
 		repository,
 		issues: [parentIssue],
 	};
 	void sourceStub;
+	// The approved final Parent body fixes the Slices checklist, so every Child
+	// title is numbered from the Initiative order rather than from this batch.
+	const parentBody = createInitiativeBody(sourceStub.repository, parent, historicalSlices);
+	const oversizedParent = bodyLimitFailure("create-initiative", parentBody);
+	if (oversizedParent) return `${oversizedParent.status}: ${oversizedParent.message}`;
+	const pendingContent = new Map<string, { title: string; body: string }>();
 	for (const operation of prepared.order) {
 		const body = childBody(repository, operation, parentIssue);
 		const oversized = bodyLimitFailure("upsert-task", body, MAX_TERMINAL_SUFFIX_BYTES);
 		if (oversized) return `${oversized.status}: ${oversized.message}`;
 		pendingContent.set(operation.task_id, {
-			title: issueTitle(`${operation.initiative_id}/${operation.slice_id}`, operation.projection?.result ?? operation.goal),
+			title: taskIssueTitle(operation, sliceOrdinalFromChecklist(parentBody, operation.slice_id, operation.projection?.slice_ordinal ?? 1)),
 			body,
 		});
 	}
-	const parentBody = createInitiativeBody(sourceStub.repository, parent, historicalSlices);
-	const oversizedParent = bodyLimitFailure("create-initiative", parentBody);
-	if (oversizedParent) return `${oversizedParent.status}: ${oversizedParent.message}`;
 	return {
 		pendingContent,
 		parent: {
-			title: issueTitle(parent.initiative_id, parent.projection?.result ?? parent.goal),
+			title: initiativeIssueTitle(parent.initiative_id, parent.projection),
 			body: parentBody,
 		},
 		parentIssueNumber: parentIssue.number,
@@ -939,20 +1093,29 @@ async function amendInitiativeParent(
 		return result(operation.op, "ambiguous_remote_state", "an amendment requires the Initiative Parent to remain open", found.issue);
 	if (binding.issue_number !== found.issue.number)
 		return result(operation.op, "ambiguous_remote_state", `amendment Parent is bound to Issue #${binding.issue_number} but observed Issue #${found.issue.number}`, found.issue);
-	if (found.issue.body === body && found.issue.title === title)
+	if (found.issue.body === body && found.issue.title === title && !labelMutationArgs(found.issue.labels, []).length)
 		return result(operation.op, "already_current", "Initiative Issue already carries the requested amended content", found.issue);
-	if (found.issue.body !== binding.body || found.issue.title !== binding.title)
+	// Content that already matches the approved final bytes still converges labels;
+	// anything else must still be the approved amendment baseline.
+	if ((found.issue.body !== body || found.issue.title !== title)
+		&& (found.issue.body !== binding.body || found.issue.title !== binding.title))
 		return result(operation.op, "ambiguous_remote_state", "Initiative Parent changed since the approved amendment baseline", found.issue);
+	// The Parent carries no Task state labels: any managed label observed on it is
+	// drift and is removed by the same edit that writes the amended content.
+	const labelArgs = labelMutationArgs(found.issue.labels, []);
 	const edited = await gh.run([
 		"issue", "edit", String(found.issue.number), "--repo", source.repository.name_with_owner,
 		"--title", title,
 		"--body-file", "-",
+		...labelArgs,
 	], { cwd: root, stdin: body });
 	if (edited.exit_code !== 0 || edited.output_exceeded) return ghFailure(operation.op, edited, "Initiative amendment edit failed");
 	const refreshed = await snapshot(root, gh, operation.op);
 	if ("contract" in refreshed) return refreshed;
 	const confirmed = lookup(refreshed.issues);
 	if (confirmed.kind !== "found") return result(operation.op, "ambiguous_remote_state", "Initiative Parent became ambiguous after amendment", found.issue);
+	if (MANAGED_ISSUE_LABELS.some((label) => confirmed.issue.labels.includes(label)))
+		return result(operation.op, "retryable_failure", "Initiative amendment did not converge to an unlabeled Parent", confirmed.issue);
 	return confirmed.issue.body === body && confirmed.issue.title === title
 		? result(operation.op, "updated", "Initiative Issue updated with approved amendment content", confirmed.issue)
 		: result(operation.op, "retryable_failure", "Initiative amendment did not converge to the requested title and body", confirmed.issue);
@@ -966,6 +1129,8 @@ async function upsertTask(
 	pendingBinding: InitiativeAmendmentBinding | undefined | null = null,
 	amendmentContext: AmendmentExecutionContext | undefined = undefined,
 ): Promise<GithubTrackerResult> {
+	const labelFailure = await labelAvailabilityFailure(root, gh, source.repository, desiredTaskLabels(operation));
+	if (labelFailure) return labelFailure;
 	const parent = initiativeLookup(source.issues, source.repository.id, operation.initiative_id);
 	if (parent.kind === "ambiguous") return result(operation.op, "ambiguous_remote_state", parent.message);
 	if (parent.kind === "missing")
@@ -997,9 +1162,10 @@ async function upsertTask(
 	const body = childBody(source.repository, operation, parent.issue);
 	const oversized = bodyLimitFailure(operation.op, body, MAX_TERMINAL_SUFFIX_BYTES);
 	if (oversized) return oversized;
-	const title = issueTitle(`${operation.initiative_id}/${operation.slice_id}`, operation.projection?.result ?? operation.goal);
+	const title = taskIssueTitle(operation, sliceOrdinalFromChecklist(parent.issue.body, operation.slice_id, operation.projection?.slice_ordinal ?? 1));
 	let child: GithubIssue;
 	let createdChild = false;
+	let labelsConverged = false;
 	if (found.kind === "missing") {
 		// Pre-create re-read (amendment path): a concurrent writer may have already
 		// created this unbound Task between the initial snapshot and our create — the
@@ -1039,6 +1205,7 @@ async function upsertTask(
 			"issue", "create", "--repo", source.repository.name_with_owner,
 			"--title", title,
 			"--body-file", "-",
+			...desiredTaskLabels(operation).flatMap((label) => ["--label", label]),
 		], { cwd: root, stdin: body });
 		const refreshed = await snapshot(root, gh, operation.op);
 		if ("contract" in refreshed) return refreshed;
@@ -1106,6 +1273,18 @@ async function upsertTask(
 		}
 		if (found.issue.body !== body || found.issue.title !== title)
 			return result(operation.op, "permanent_failure", "Task Issue already exists with a different title or Agent Brief; edit the GitHub source or retry the original projection before changing native relations", found.issue);
+		// A repeated complete batch is a convergence pass: managed label drift is
+		// repaired here without rewriting content the tracker never owns.
+		const observedLabelArgs = labelMutationArgs(child.labels, desiredTaskLabels(operation));
+		if (observedLabelArgs.length) {
+			const edited = await gh.run([
+				"issue", "edit", String(child.number), "--repo", source.repository.name_with_owner,
+				...observedLabelArgs,
+			], { cwd: root });
+			if (edited.exit_code !== 0 || edited.output_exceeded)
+				return ghFailure(operation.op, edited, `Task Issue #${child.number} label convergence failed`);
+			labelsConverged = true;
+		}
 	}
 	const attachment = await confirmAttachment(root, gh, operation.op, source.repository, parent.issue.number, child.number);
 	if (!("attached" in attachment)) return attachment;
@@ -1141,7 +1320,7 @@ async function upsertTask(
 	if (attachment.attached && dependencies.complete)
 		return createdChild
 			? result(operation.op, "created", "Task Issue created and attached with native blocking relations", child)
-			: result(operation.op, "already_current", "Task Issue, native Sub-issue relation, and blocking relations are current", child);
+			: result(operation.op, labelsConverged ? "updated" : "already_current", "Task Issue, native Sub-issue relation, and blocking relations are current", child);
 	return createdChild
 		? result(operation.op, "created", "Task Issue created and attached as a native Sub-issue", child)
 		: result(operation.op, "updated", "existing Task Issue attached as a native Sub-issue", child);
@@ -1272,7 +1451,13 @@ function normalizeProjection(value: TaskProjection | undefined): TaskProjection
 	if (!value || typeof value !== "object" || Array.isArray(value)) throw new Error("projection must be an object");
 	const blockedBy = normalizedProjectionList(value.blocked_by, "projection.blocked_by", 128)?.map((id) => identifier(id, "projection.blocked_by task_id"));
 	if (blockedBy && new Set(blockedBy).size !== blockedBy.length) throw new Error("projection.blocked_by must not contain duplicate Task IDs");
+	if (value.slice_ordinal !== undefined
+		&& (typeof value.slice_ordinal !== "number" || !Number.isSafeInteger(value.slice_ordinal) || value.slice_ordinal < 1 || value.slice_ordinal > 999))
+		throw new Error("projection.slice_ordinal must be an integer between 1 and 999");
 	return {
+		short_name: value.short_name,
+		title: value.title,
+		slice_ordinal: value.slice_ordinal,
 		result: value.result === undefined ? undefined : projectionText(value.result, "projection.result"),
 		current_behavior: value.current_behavior === undefined ? undefined : projectionText(value.current_behavior, "projection.current_behavior"),
 		desired_behavior: value.desired_behavior === undefined ? undefined : projectionText(value.desired_behavior, "projection.desired_behavior"),
@@ -1287,7 +1472,12 @@ function normalizeProjection(value: TaskProjection | undefined): TaskProjection
 function normalizeInitiativeProjection(value: InitiativeProjection | undefined): InitiativeProjection | undefined {
 	if (value === undefined) return undefined;
 	if (!value || typeof value !== "object" || Array.isArray(value)) throw new Error("projection must be an object");
+	if (value.source_issue !== undefined && (typeof value.source_issue !== "string" || !/^[1-9][0-9]{0,9}$/.test(value.source_issue)))
+		throw new Error("projection.source_issue must be a GitHub Issue number");
 	return {
+		short_name: value.short_name,
+		title: value.title,
+		source_issue: value.source_issue,
 		problem: value.problem === undefined ? undefined : projectionText(value.problem, "projection.problem"),
 		result: value.result === undefined ? undefined : projectionText(value.result, "projection.result"),
 		design: value.design === undefined ? undefined : projectionText(value.design, "projection.design"),
@@ -1318,7 +1508,7 @@ function validateOperation(operation: TrackerOperation): TrackerOperation {
 			};
 			}),
 		};
-		issueTitle(normalized.initiative_id, normalized.projection?.result ?? normalized.goal);
+		initiativeIssueTitle(normalized.initiative_id, normalized.projection);
 		return normalized;
 	}
 	if (operation.op === "upsert-task") {
@@ -1335,7 +1525,7 @@ function validateOperation(operation: TrackerOperation): TrackerOperation {
 				summary: projectionText(item.summary, `acceptance[${index}].summary`, 500),
 			})),
 		};
-		issueTitle(`${normalized.initiative_id}/${normalized.slice_id}`, normalized.projection?.result ?? normalized.goal);
+		taskIssueTitle(normalized);
 		return normalized;
 	}
 	return {
@@ -1446,10 +1636,14 @@ function preflightPublication(root: string, input: InitiativePublicationInput):
 		if (input.projection[field] === undefined)
 			throw new Error(`a complete Initiative publication requires projection.${field}`);
 	}
+	// The Initiative display names are validated once here, before any Task
+	// projection is stamped with them, so a missing or oversized display name
+	// fails the batch closed with zero remote writes.
+	const initiativeShortName = initiativeDisplayNames(input.projection).shortName;
 	const publications = input.tasks.map((task, index) => {
 		if (!task || typeof task !== "object" || Array.isArray(task)) throw new Error(`tasks[${index}] must be an object`);
 		if (typeof task.intent !== "string") throw new Error(`tasks[${index}].intent must be a string`);
-		return taskPublication(root, input.initiative_id, task.slice_id, task.intent, task.acceptance, task.projection);
+		return taskPublication(root, input.initiative_id, task.slice_id, task.intent, task.acceptance, task.projection, index + 1, initiativeShortName);
 	});
 	const historicalIds = new Set<string>();
 	if (input.amendment) {
@@ -1820,7 +2014,7 @@ async function updatePendingChild(
 	const body = childBody(source.repository, op, parent.kind === "found" ? parent.issue : child);
 	const oversized = bodyLimitFailure(op.op, body, MAX_TERMINAL_SUFFIX_BYTES);
 	if (oversized) return oversized;
-	const title = issueTitle(`${op.initiative_id}/${op.slice_id}`, op.projection?.result ?? op.goal);
+	const title = taskIssueTitle(op, sliceOrdinalFromChecklist(parent.kind === "found" ? parent.issue.body : child.body, op.slice_id, op.projection?.slice_ordinal ?? 1));
 	// Parent content expectation for pre-write revalidation: on the amendment
 	// path the expectation is always the fixed approved final Parent bytes —
 	// never re-adopt freshly observed content as the baseline, so an edit that
@@ -1874,7 +2068,8 @@ async function updatePendingChild(
 	// that landed after the snapshot must be preserved exactly, never
 	// overwritten by bytes computed from stale data.
 	let observed = child;
-	if (child.title !== title || child.body !== finalBody) {
+	const desiredLabels = desiredTaskLabels(op);
+	if (child.title !== title || child.body !== finalBody || labelMutationArgs(child.labels, desiredLabels).length) {
 		// Re-read the Child immediately before writing: earlier blocker ownership
 		// reads may have raced a concurrent user edit. The bound issue_number must
 		// still hold and the remote content must still be baseline-or-approved-final.
@@ -1896,11 +2091,13 @@ async function updatePendingChild(
 		const writeBody = typeof observedEvent === "string"
 			? `${body.trimEnd()}${terminalSuffix(observedEvent)}`
 			: finalBody;
-		if (observed.title !== title || observed.body !== writeBody) {
+		const labelArgs = labelMutationArgs(observed.labels, desiredLabels);
+		if (observed.title !== title || observed.body !== writeBody || labelArgs.length) {
 			const edited = await gh.run([
 				"issue", "edit", String(child.number), "--repo", source.repository.name_with_owner,
 				"--title", title,
 				"--body-file", "-",
+				...labelArgs,
 			], { cwd: root, stdin: writeBody });
 			if (edited.exit_code !== 0 || edited.output_exceeded) return ghFailure(op.op, edited, `pending Task Issue #${child.number} update failed`);
 			finalBody = writeBody;
@@ -1915,11 +2112,15 @@ async function updatePendingChild(
 		return result(op.op, "ambiguous_remote_state", `pending Task ${op.task_id} changed identity during amendment`, child);
 	if (reread.issue.title !== title || reread.issue.body !== finalBody)
 		return result(op.op, "retryable_failure", `pending Task ${op.task_id} update did not converge`, reread.issue);
+	const labelsCurrent = !desiredLabels.some((label) => !reread.issue.labels.includes(label));
+	if (!labelsCurrent)
+		return result(op.op, "retryable_failure", `pending Task ${op.task_id} labels did not converge`, reread.issue);
 	const currentDependencies = await confirmBlockedBy(root, gh, op.op, refreshed.repository, reread.issue.number, desiredBlockers);
 	if (!("complete" in currentDependencies)) return currentDependencies;
 	if (!currentDependencies.complete)
 		return result(op.op, "retryable_failure", `pending Task ${op.task_id} dependencies did not converge`, reread.issue);
-	const contentCurrent = child.title === title && child.body === finalBody;
+	const contentCurrent = child.title === title && child.body === finalBody
+		&& labelMutationArgs(child.labels, desiredLabels).length === 0;
 	return contentCurrent
 		? result(op.op, "already_current", `pending Task ${op.task_id} already carries the approved amendment content`, reread.issue)
 		: result(op.op, "updated", `pending Task ${op.task_id} Agent Brief updated with approved amendment content`, reread.issue);
@@ -2089,6 +2290,12 @@ export async function runGithubInitiativePublication(
 	if (initialParent.kind === "ambiguous") return publicationResult("ambiguous_remote_state", initialParent.message);
 	if (amendment && initialParent.kind === "missing")
 		return publicationResult("permanent_failure", "an amendment requires the Initiative Parent to already exist");
+	// Label availability is validated before any remote mutation: the tracker
+	// never creates labels, so a missing label fails the whole batch closed
+	// instead of half-publishing an Initiative.
+	const requiredLabels = [...new Set(prepared.order.flatMap((operation) => desiredTaskLabels(operation)))];
+	const labelFailure = await labelAvailabilityFailure(absoluteRoot, gh, initial.repository, requiredLabels);
+	if (labelFailure) return publicationResult(labelFailure.status, labelFailure.message, labelFailure);
 	const parentForPreflight = initialParent.kind === "found" ? initialParent.issue : {
 		id: Number.MAX_SAFE_INTEGER,
 		number: Number.MAX_SAFE_INTEGER,
@@ -2097,6 +2304,7 @@ export async function runGithubInitiativePublication(
 		body: "",
 		state: "open" as const,
 		state_reason: null,
+		labels: [],
 	};
 	const parentBodyFailure = bodyLimitFailure("create-initiative", createInitiativeBody(initial.repository, prepared.initiative));
 	if (parentBodyFailure) return publicationResult(parentBodyFailure.status, parentBodyFailure.message, parentBodyFailure);
@@ -2214,7 +2422,7 @@ export async function runGithubInitiativePublication(
 	const parentDrift = publicationIssueDrift(
 		parentResult,
 		parent.issue,
-		amendmentContext ? amendmentContext.parent.title : issueTitle(prepared.initiative.initiative_id, prepared.initiative.projection?.result ?? prepared.initiative.goal),
+		amendmentContext ? amendmentContext.parent.title : initiativeIssueTitle(prepared.initiative.initiative_id, prepared.initiative.projection),
 		amendmentContext
 			? amendmentContext.parent.body
 			: createInitiativeBody(finalSource.repository, prepared.initiative),
@@ -2241,12 +2449,15 @@ export async function runGithubInitiativePublication(
 		const childDrift = publicationIssueDrift(
 			childResult,
 			child.issue,
-			issueTitle(`${operation.initiative_id}/${operation.slice_id}`, operation.projection?.result ?? operation.goal),
+			taskIssueTitle(operation, sliceOrdinalFromChecklist(parent.issue.body, operation.slice_id, operation.projection?.slice_ordinal ?? 1)),
 			childBody(finalSource.repository, operation, parent.issue),
 			`Task ${operation.task_id}`,
 			{ allowTerminalSuffix: amendmentContext !== undefined },
 		);
 		if (childDrift) return publicationResult("ambiguous_remote_state", childDrift, parentResult, taskResults);
+		const expectedLabels = desiredTaskLabels(operation);
+		if (childResult.status !== "already_current" && expectedLabels.some((label) => !child.issue.labels.includes(label)))
+			return publicationResult("ambiguous_remote_state", `Task ${operation.task_id} is missing publication labels after publication`, parentResult, taskResults);
 		expectedNumbers.push(child.issue.number);
 		const ownership = await confirmTerminalOwnership(absoluteRoot, gh, "upsert-task", finalSource, child.issue);
 		if (!("owned" in ownership)) return publicationResult(ownership.status, ownership.message, parentResult, taskResults);
@@ -2341,7 +2552,9 @@ function taskPublication(
 	sliceId: string,
 	intentPath: string,
 	acceptance: unknown,
-	projection?: TaskProjection,
+	projection: TaskProjection | undefined,
+	ordinal: number,
+	initiativeShortName: string,
 ): PreparedPublicationTask {
 	const absoluteRoot = resolve(root);
 	const absolutePath = resolve(absoluteRoot, intentPath);
@@ -2367,6 +2580,18 @@ function taskPublication(
 	});
 	const missingIds = [...expectedIds].filter((id) => !publicById.has(id));
 	if (missingIds.length) throw new Error(`Task ${taskId} is missing public acceptance ids: ${missingIds.join(", ")}`);
+	// Display names are stamped from the Initiative so every Child title carries
+	// the same short name and its declared Slice position; an explicitly
+	// supplied value must agree, never silently override.
+	if (projection?.short_name !== undefined && projection.short_name !== initiativeShortName)
+		throw new Error(`Task ${taskId} projection.short_name must match the Initiative short name`);
+	if (projection?.slice_ordinal !== undefined && projection.slice_ordinal !== ordinal)
+		throw new Error(`Task ${taskId} projection.slice_ordinal must match its declared Slice position ${ordinal}`);
+	const stampedProjection: TaskProjection = {
+		...(projection ?? {}),
+		short_name: initiativeShortName,
+		slice_ordinal: ordinal,
+	};
 	return {
 		operation: validateOperation({
 			op: "upsert-task",
@@ -2376,7 +2601,7 @@ function taskPublication(
 			goal: intent.goal,
 			risk: intent.risk,
 			acceptance: intent.acceptance.map((item) => publicById.get(item.id)!),
-			projection,
+			projection: stampedProjection,
 		}) as Extract<TrackerOperation, { op: "upsert-task" }>,
 		intent_path: read.intent_ref.path,
 		intent_content_hash: read.content_hash,

package/plugins/immune-brain/runtime/plugin_version.ts

@@ -1,2 +1,2 @@
 // Generated by scripts/plugin_versioning.ts from the root package.json.
-export const PLUGIN_VERSION = "3.6.8" as const;
+export const PLUGIN_VERSION = "3.6.9" as const;

package/plugins/immune-brain/skills/imm-review-retro/SKILL.md

@@ -0,0 +1,23 @@
+---
+name: imm-review-retro
+description: Use when the user explicitly requests Immune-Brain ranking of models by cross-model review load or a project usage retro.
+---
+
+# Immune-Brain: Review Retro
+
+Use [`../../dist/imm-review-retro.md`](../../dist/imm-review-retro.md) as the
+canonical contract index, not a whole-document read. This is a
+standalone host-native analysis entry, not a Managed Path continuation
+and not an `imm-loop` internal-role dispatch.
+
+Mandatory constraints: read-only on pi session logs. Do not edit code, tests,
+Specs, or workflow state. Do not write session logs or `.imm/` files.
+
+Section routes - load a section's instructions only when its branch applies.
+Read each linked heading body up to the next heading; nested sections and
+references load only under their own condition. Never read the whole contract
+or all references as an entry prerequisite.
+
+- common: [Boundary](../../dist/imm-review-retro.md#boundary), [Invocation](../../dist/imm-review-retro.md#invocation)
+- running the analyzer: [Counting rules](../../dist/imm-review-retro.md#counting-rules), [CLI](../../dist/imm-review-retro.md#cli)
+- interpreting the report: [Report](../../dist/imm-review-retro.md#report), [Caveats](../../dist/imm-review-retro.md#caveats)

package/plugins/immune-brain/skills/imm-review-retro/scripts/review_retro.ts

@@ -0,0 +1,355 @@
+#!/usr/bin/env bun
+/**
+ * Retro: how many code reviews did each model's own code trigger, across recent pi sessions.
+ *
+ * Usage: bun review_retro.ts <days> [--root <sessions-dir>] [--project <substr>] [--top N]
+ *
+ * Counting rules (the 口径 that keeps the numbers honest):
+ *   review executed      = Agent(subagent_type="Review") tool call
+ *   avgSc / pass%        = average score (0-10) and PASS rate from [SCORE: ...] tags in Review toolResult
+ *   kernel:submit_review = the *registration* of that same review, reported separately (never added)
+ *   attribution          = the model behind the most recent edit/write before the review (the code's author)
+ *   findings             = imm_kernel_canary record_finding, deduped per session, harness bookkeeping split out
+ */
+import { createReadStream, existsSync, readdirSync } from "node:fs";
+import { homedir } from "node:os";
+import { join } from "node:path";
+import { createInterface } from "node:readline";
+import { pathToFileURL } from "node:url";
+
+const BOOKKEEPING = /recorded cleanly|receipt recorded|round recorded|no findings?\b/i;
+const EDIT_TOOLS = new Set(["edit", "write", "multiedit"]);
+const SCORE_RE = /\[SCORE:\s*([\d.]+)\s*(?:\/\s*10)?\]/i;
+const VERDICT_RE = /\[VERDICT:\s*(\w+)\]/i;
+const RISK_RE = /\[RISK:\s*(\w+)\]/i;
+const BLOCK_RE = /\[BLOCKING:\s*(\d+)\]/i;
+const ADVIS_RE = /\[ADVISORY:\s*(\d+)\]/i;
+
+type ReviewTag = {
+	score: number;
+	verdict: string;
+	risk: string;
+	blocking: number;
+	advisory: number;
+};
+
+function parseReviewTag(text: string): ReviewTag | null {
+	if (!text.includes("[SCORE:")) return null;
+	const sc = SCORE_RE.exec(text);
+	if (!sc) return null;
+	const score = Number(sc[1]);
+	if (!Number.isFinite(score)) return null;
+	return {
+		score,
+		verdict: (VERDICT_RE.exec(text)?.[1] ?? "UNKNOWN").toUpperCase(),
+		risk: (RISK_RE.exec(text)?.[1] ?? "UNKNOWN").toUpperCase(),
+		blocking: Number(BLOCK_RE.exec(text)?.[1] ?? 0),
+		advisory: Number(ADVIS_RE.exec(text)?.[1] ?? 0),
+	};
+}
+
+function extractText(content: unknown): string {
+	if (typeof content === "string") return content;
+	if (Array.isArray(content)) {
+		return content
+			.map((item) => {
+				if (typeof item === "string") return item;
+				if (item && typeof item === "object") {
+					const rec = item as Record<string, unknown>;
+					if ("text" in rec) return String(rec.text);
+					if ("result" in rec) return String(rec.result);
+				}
+				return "";
+			})
+			.join("\n");
+	}
+	return "";
+}
+
+function norm(a: unknown): Record<string, unknown> {
+	if (typeof a === "string") {
+		try {
+			return JSON.parse(a) as Record<string, unknown>;
+		} catch {
+			return {};
+		}
+	}
+	return a && typeof a === "object" ? (a as Record<string, unknown>) : {};
+}
+
+function bump(map: Map<string, number>, key: string, n = 1): void {
+	map.set(key, (map.get(key) ?? 0) + n);
+}
+
+type Args = { days: number; root: string; project: string; top: number };
+
+function parseArgs(argv: string[]): Args {
+	const out: Args = {
+		days: NaN,
+		root: join(homedir(), ".pi/agent/sessions"),
+		project: "",
+		top: 15,
+	};
+	const rest: string[] = [];
+	for (let i = 0; i < argv.length; i++) {
+		const a = argv[i]!;
+		if (a === "--root") out.root = argv[++i] ?? "";
+		else if (a === "--project") out.project = argv[++i] ?? "";
+		else if (a === "--top") out.top = Number(argv[++i]);
+		else if (a.startsWith("-")) throw new Error(`unknown flag: ${a}`);
+		else rest.push(a);
+	}
+	out.days = Number(rest[0]);
+	if (!(out.days > 0) || rest[1] !== undefined) {
+		throw new Error("usage: review_retro.ts <days> [--root <dir>] [--project <substr>] [--top N]");
+	}
+	if (!Number.isFinite(out.top) || out.top <= 0) throw new Error("--top must be > 0");
+	return out;
+}
+
+function listJsonl(root: string): string[] {
+	if (!existsSync(root)) return [];
+	const paths: string[] = [];
+	for (const dir of readdirSync(root, { withFileTypes: true })) {
+		if (!dir.isDirectory()) continue;
+		const folder = join(root, dir.name);
+		for (const name of readdirSync(folder)) {
+			if (name.endsWith(".jsonl")) paths.push(join(folder, name));
+		}
+	}
+	return paths;
+}
+
+export async function run(argv: string[]): Promise<string> {
+	const args = parseArgs(argv);
+	const cut = new Date(Date.now() - args.days * 86400000).toISOString().slice(0, 19);
+	const home = homedir();
+	const dev = new Map<string, number>();
+	const turns = new Map<string, number>();
+	const rev = new Map<string, number>();
+	const sub = new Map<string, number>();
+	const tools = new Map<string, number>();
+	const proj = new Map<string, number>();
+	const episode = new Map<string, Set<string>>();
+	const tasks = new Map<string, Set<string>>();
+	const findUniq = new Map<string, Set<string>>();
+	const findingsRaw = new Map<string, number>();
+	const scores = new Map<string, number[]>();
+	const verdicts = new Map<string, number>();
+	const risks = new Map<string, number>();
+	let files = 0;
+
+	for (const path of listJsonl(args.root)) {
+		let editor: string | null = null;
+		let cwd = "";
+		let used = false;
+		const pending = new Map<string, string>();
+		const rl = createInterface({ input: createReadStream(path, { encoding: "utf8" }) });
+		for await (const raw of rl) {
+			const line = raw.trim();
+			if (!line) continue;
+			let o: Record<string, unknown>;
+			try {
+				o = JSON.parse(line) as Record<string, unknown>;
+			} catch {
+				continue;
+			}
+			if (o.type === "session") {
+				cwd = String(o.cwd ?? "");
+				continue;
+			}
+			const m = (o.message && typeof o.message === "object" ? o.message : {}) as Record<string, unknown>;
+			const role = m.role;
+			if (String(o.timestamp ?? "") < cut) continue;
+			if (args.project && !cwd.includes(args.project)) continue;
+
+			if (role === "assistant") {
+				used = true;
+				const mo = `${m.provider ?? "?"}/${m.model ?? "?"}`;
+				bump(turns, mo);
+				const content = Array.isArray(m.content) ? m.content : [];
+				for (const c of content) {
+					if (!c || typeof c !== "object") continue;
+					const call = c as Record<string, unknown>;
+					if (call.type !== "toolCall") continue;
+					const name = String(call.name ?? "");
+					const a = norm(call.arguments);
+					const cid = typeof call.id === "string" ? call.id : "";
+					bump(tools, name || "?");
+					if (EDIT_TOOLS.has(name)) {
+						bump(dev, mo);
+						editor = mo;
+					} else if (name === "Agent" && a.subagent_type === "Review") {
+						const owner = editor ?? "no-edit (review-only)";
+						bump(rev, owner);
+						const ep = episode.get(owner) ?? new Set();
+						ep.add(`${path}\0${String(a.description ?? "")}${String(a.prompt ?? "").slice(0, 240)}`);
+						episode.set(owner, ep);
+						const projKey = `${cwd.replace(home, "~")}\0${owner}`;
+						bump(proj, projKey);
+						if (cid) pending.set(cid, owner);
+					} else if (name === "imm_kernel_canary") {
+						const act = norm(a.action);
+						const op = act.op;
+						const owner = editor ?? "no-edit (review-only)";
+						if (op === "submit_review") {
+							bump(sub, owner);
+							const t = tasks.get(owner) ?? new Set();
+							t.add(`${cwd}\0${String(a.task_id ?? "")}`);
+							tasks.set(owner, t);
+						} else if (op === "record_finding") {
+							const f = norm(act.finding);
+							const summ = String(f.summary ?? "");
+							const kind = BOOKKEEPING.test(summ) ? "bookkeeping" : String(f.kind ?? "");
+							const rawKey = `${owner}\0${kind}`;
+							bump(findingsRaw, rawKey);
+							const uniq = findUniq.get(rawKey) ?? new Set();
+							uniq.add(`${path}\0${summ.slice(0, 160)}`);
+							findUniq.set(rawKey, uniq);
+						}
+					}
+				}
+			} else if (role === "toolResult") {
+				const tcid = String(m.toolCallId ?? "");
+				const owner = pending.get(tcid);
+				if (!owner) continue;
+				pending.delete(tcid);
+				const tag = parseReviewTag(extractText(m.content));
+				if (!tag) continue;
+				const sl = scores.get(owner) ?? [];
+				sl.push(tag.score);
+				scores.set(owner, sl);
+				bump(verdicts, `${owner}\0${tag.verdict}`);
+				bump(risks, `${owner}\0${tag.risk}`);
+			}
+		}
+		if (used) files += 1;
+	}
+
+	const lines: string[] = [];
+	const emit = (s = "") => lines.push(s);
+	emit(`window: last ${args.days}d (UTC >= ${cut}Z) | sessions with activity: ${files} | root: ${args.root}`);
+	emit("review = Agent(Review) executed; attributed to the model that last edited the code under review");
+	emit("avgSc/pass% = parsed from Review [SCORE: .../10] [VERDICT: ...] tags (shows '-' if untagged)");
+	emit("");
+	const hdr =
+		`${"model".padEnd(42)}${"devEdits".padStart(9)}${"turns".padStart(6)}${"reviews".padStart(8)}${"uniq".padStart(5)}${"rev/100ed".padStart(10)}${"avgSc".padStart(6)}${"pass%".padStart(6)}${"registr".padStart(8)}${"block".padStart(6)}${"advis".padStart(6)}${"noisy".padStart(6)}`;
+	emit(hdr);
+	emit("-".repeat(hdr.length));
+
+	const models = [...new Set([...rev.keys(), ...dev.keys()])].sort(
+		(a, b) => (rev.get(b) ?? 0) - (rev.get(a) ?? 0),
+	);
+	const shown: string[] = [];
+	for (const mo of models) {
+		if (!rev.get(mo) && (dev.get(mo) ?? 0) < 30) continue;
+		shown.push(mo);
+		const uq = episode.get(mo)?.size ?? 0;
+		const d = dev.get(mo) ?? 0;
+		const r = rev.get(mo) ?? 0;
+		const rate = d ? ((100 * r) / d).toFixed(1) : "-";
+		const sl = scores.get(mo) ?? [];
+		const avgSc = sl.length ? (sl.reduce((x, y) => x + y, 0) / sl.length).toFixed(1) : "-";
+		const passCnt = verdicts.get(`${mo}\0PASS`) ?? 0;
+		const passPct = sl.length ? `${Math.round((100 * passCnt) / sl.length)}%` : "-";
+		const block = findUniq.get(`${mo}\0blocking`)?.size ?? 0;
+		const advis = findUniq.get(`${mo}\0advisory`)?.size ?? 0;
+		const noisy = findUniq.get(`${mo}\0bookkeeping`)?.size ?? 0;
+		emit(
+			`${mo.padEnd(42)}${String(d).padStart(9)}${String(turns.get(mo) ?? 0).padStart(6)}${String(r).padStart(8)}${String(uq).padStart(5)}${rate.padStart(10)}${avgSc.padStart(6)}${passPct.padStart(6)}${String(sub.get(mo) ?? 0).padStart(8)}${String(block).padStart(6)}${String(advis).padStart(6)}${String(noisy).padStart(6)}`,
+		);
+	}
+	emit("-".repeat(hdr.length));
+	const totalScores = [...scores.values()].flat();
+	const totAvg = totalScores.length
+		? (totalScores.reduce((x, y) => x + y, 0) / totalScores.length).toFixed(1)
+		: "-";
+	const totPass = shown.reduce((n, mo) => n + (verdicts.get(`${mo}\0PASS`) ?? 0), 0);
+	const totPassPct = totalScores.length ? `${Math.round((100 * totPass) / totalScores.length)}%` : "-";
+	const totBlock = [...findUniq.entries()].reduce((n, [k, v]) => n + (k.endsWith("\0blocking") ? v.size : 0), 0);
+	const totAdvis = [...findUniq.entries()].reduce((n, [k, v]) => n + (k.endsWith("\0advisory") ? v.size : 0), 0);
+	const totNoisy = [...findUniq.entries()].reduce((n, [k, v]) => n + (k.endsWith("\0bookkeeping") ? v.size : 0), 0);
+	const totDev = [...dev.values()].reduce((a, b) => a + b, 0);
+	const totTurns = [...turns.values()].reduce((a, b) => a + b, 0);
+	const totRev = [...rev.values()].reduce((a, b) => a + b, 0);
+	const totUniq = [...episode.values()].reduce((n, s) => n + s.size, 0);
+	const totSub = [...sub.values()].reduce((a, b) => a + b, 0);
+	emit(
+		`${"TOTAL".padEnd(42)}${String(totDev).padStart(9)}${String(totTurns).padStart(6)}${String(totRev).padStart(8)}${String(totUniq).padStart(5)}${"".padStart(10)}${totAvg.padStart(6)}${totPassPct.padStart(6)}${String(totSub).padStart(8)}${String(totBlock).padStart(6)}${String(totAdvis).padStart(6)}${String(totNoisy).padStart(6)}`,
+	);
+
+	emit("");
+	emit("=== usage (sessions / turns / edits / tools) ===");
+	emit(`sessions: ${files} | turns: ${totTurns} | edits: ${totDev}`);
+	const toolRows = [...tools.entries()].sort((a, b) => b[1] - a[1]);
+	if (toolRows.length === 0) emit("(no tool calls in this window)");
+	else for (const [name, n] of toolRows) emit(`${String(n).padStart(5)}  ${name}`);
+
+	emit("");
+	emit("=== review quality & scores (new rubric) ===");
+	const scoredModels = [...scores.keys()]
+		.filter((m) => (scores.get(m) ?? []).length > 0)
+		.sort((a, b) => (scores.get(b)?.length ?? 0) - (scores.get(a)?.length ?? 0));
+	if (scoredModels.length) {
+		const qHdr = `${"model".padEnd(42)}${"scored".padStart(7)}${"avgScore".padStart(9)}${"PASS".padStart(6)}${"REVISE".padStart(7)}${"REJECT".padStart(7)}${"highRisk".padStart(9)}`;
+		emit(qHdr);
+		emit("-".repeat(qHdr.length));
+		for (const mo of scoredModels) {
+			const sl = scores.get(mo) ?? [];
+			const avgS = (sl.reduce((x, y) => x + y, 0) / sl.length).toFixed(2);
+			const pC = verdicts.get(`${mo}\0PASS`) ?? 0;
+			const revC = verdicts.get(`${mo}\0REVISE`) ?? 0;
+			const rejC = verdicts.get(`${mo}\0REJECT`) ?? 0;
+			const highR = (risks.get(`${mo}\0HIGH`) ?? 0) + (risks.get(`${mo}\0CRITICAL`) ?? 0);
+			emit(
+				`${mo.padEnd(42)}${String(sl.length).padStart(7)}${avgS.padStart(9)}${String(pC).padStart(6)}${String(revC).padStart(7)}${String(rejC).padStart(7)}${String(highR).padStart(9)}`,
+			);
+		}
+	} else {
+		emit("(no scored reviews found in this window yet; reviews with [SCORE: .../10] will appear here)");
+	}
+
+	emit("");
+	emit("=== where the reviews landed (project x author) ===");
+	const projRows = [...proj.entries()].sort((a, b) => b[1] - a[1]).slice(0, args.top);
+	for (const [key, v] of projRows) {
+		const [p, mo] = key.split("\0");
+		emit(`${String(v).padStart(5)}  ${String(p).padEnd(58)} ${mo}`);
+	}
+
+	emit("");
+	emit("=== review rounds per kernel task (rework signal) ===");
+	const taskModels = [...tasks.keys()].sort((a, b) => (sub.get(b) ?? 0) - (sub.get(a) ?? 0));
+	for (const mo of taskModels) {
+		const nt = tasks.get(mo)?.size ?? 0;
+		if (!nt) continue;
+		const s = sub.get(mo) ?? 0;
+		emit(
+			`${(s / nt).toFixed(1).padStart(6)} rounds/task   ${String(s).padStart(4)} registrations / ${String(nt).padStart(3)} tasks   ${mo}`,
+		);
+	}
+
+	const rawSum = [...findingsRaw.values()].reduce((a, b) => a + b, 0);
+	const uniqSum = [...findUniq.values()].reduce((n, s) => n + s.size, 0);
+	if (rawSum && rawSum !== uniqSum) {
+		emit("");
+		emit(
+			`(findings: ${rawSum} raw calls deduped to ${uniqSum} distinct — re-submits of the same finding are counted once)`,
+		);
+	}
+	emit("caveats: 'bookkeep' = kernel self-entries mislabelled as findings, excluded from blocking/advisory;");
+	emit("         high rounds/task can be canary/QA harness re-registration, not human-visible rework.");
+	return `${lines.join("\n")}\n`;
+}
+
+const isMain = Boolean(process.argv[1]) && pathToFileURL(process.argv[1]!).href === import.meta.url;
+if (isMain) {
+	run(process.argv.slice(2))
+		.then((text) => {
+			process.stdout.write(text);
+		})
+		.catch((err: unknown) => {
+			console.error(err);
+			process.exit(1);
+		});
+}

package/plugins/immune-brain/skills/registry.yaml

@@ -56,3 +56,12 @@ skills:
     output_artifacts: [maintain_report]
     next_actions: []
     boundary: Minimize tracked agent-instruction context after explicit manifest approval; no Managed authority mutation, contract installation, or reference-document creation.
+  - name: imm-review-retro
+    path: skills/imm-review-retro/SKILL.md
+    role: execute
+    title: Review Retro
+    role_class: discovery
+    canonical: true
+    output_artifacts: [retro_report]
+    next_actions: []
+    boundary: Rank models by review load and report project usage from pi session logs. Read-only. No Managed Path mutation.