@blxzer/cursor-trellis 0.1.3 → 0.2.1

package/dist/templates/trellis/workflow.md

@@ -41,8 +41,6 @@ python3 ./.trellis/scripts/get_context.py --mode packages   # list packages / la
 
 Every task has its own directory under `.trellis/tasks/{MM-DD-name}/` holding `task.json`, `prd.md`, optional `design.md`, optional `implement.md`, optional `research/`, and context manifests (`implement.jsonl`, `check.jsonl`) for Cursor (sub-agent dispatch).
 
-Optional copied templates live under `.trellis/tasks/templates/` (for example `release-readiness/` and `release-execution/`). `task.py create` does not auto-apply them; copy files into a new task when starting that workflow.
-
 ```bash
 # Task lifecycle
 python3 ./.trellis/scripts/task.py create "<title>" [--slug <name>] [--parent <dir>]
@@ -265,7 +263,11 @@ python3 ./.trellis/scripts/task.py review-child <parent-task> <child-task> --dec
 
 - `generate-child-prompt` reads parent `task-map.md` for `depends_on` and `touches`, child artifacts, and optional parent `child-prompts.md`. Use `--mode subagent` only as a delivery hint when the platform can spawn subagents; inline mode remains the portable default.
 - `review-child` summarizes child `verify.md` / `handoff.md`, appends notes to parent `verify.md`, and can advance `accepted` / `integrating` / `integrated` in one flow (`--decision integrate-through`) while still using the same Stage 0 integration guards as `integrate-child`.
-- Reviewer quality gates (`child-review`, `parent-accepted`, `parent-integrated`) are **not** auto-recorded; the command prints optional `record-gate` hints only.
+- Reviewer quality gates are **not** auto-recorded. CLI enforces them at transition boundaries:
+  - **Full Child accept / integrate-through**: requires substantive `verify.md` evidence and `child-review/code-review` (plus configured architecture gates) before Parent marks the Child `accepted`.
+  - **Parent archive**: requires every structural Child `integrated` or `cancelled`, substantive Parent integration evidence, and `parent-integrated/integration-review`.
+  - **Lite closeout**: explicit no-gate chain; archive still requires validation, acceptance, and durable-learning evidence in `verify.md`.
+  - **`record-gate`**: rejects PASS/SKIPPED when transition evidence is missing or placeholder-only.
 
 <!-- Per-turn breadcrumb: shown when no task is selected (before Phase 1) -->
 
@@ -724,10 +726,10 @@ Archive readiness by mode:
 |---|---|
 | No Task | No archive; upgrade to a durable task mode before archive is possible. |
 | Micro-Grill | No archive unless upgraded into Lite, Full, or Parent/Child. |
-| Lite | `verify.md` has validation evidence, final acceptance evidence, and durable-learning decision evidence. |
-| Full | Lite evidence plus required completion gates, fresh fingerprints, and no unresolved required `FAIL` gates. |
-| Child | Lite/Full evidence plus Parent task-map marks the Child `integrated` or `cancelled`; integrated Children include `handoff.md`. |
-| Parent | Full evidence plus every Child is `integrated` or `cancelled`, with final integration evidence. |
+| Lite | Explicit no-gate chain. `verify.md` has substantive validation, final acceptance, and durable-learning decision evidence. |
+| Full | Lite evidence plus required completion gates (`full-task-complete/*`), substantive check/change-set evidence, fresh fingerprints, and no unresolved required `FAIL` gates. |
+| Child | Lite or Full evidence (by child profile). Full Children require `child-review` gates before Parent `accepted`. Parent task-map marks the Child `integrated` or `cancelled`; integrated Children include `handoff.md`. |
+| Parent | Archive evidence plus every Child `integrated` or `cancelled`, substantive final integration evidence, and `parent-integrated/integration-review`. |
 
 Archive / Learning is terminal. After archive, do not silently mutate archived task artifacts; follow-up work requires a new task unless the user explicitly approves an archive amendment.
 
@@ -735,52 +737,6 @@ If archive is not being run in this session, report the passing or failing archi
 
 ---
 
-## Release readiness and release execution
-
-Package and CLI releases use **two tasks** (or two explicit phases with separate evidence). Readiness is non-mutating; execution requires explicit user approval before any remote mutation.
-
-### When to split
-
-| Task type | Purpose | Remote mutations |
-| --- | --- | --- |
-| **Release readiness** | Version recommendation, changelog, manifest notes, build/test/typecheck, pack dry-run, prepublish blockers | **None** — no publish, tag, push, or version bump |
-| **Release execution** | Preflight immediately before publish; run approved `release.js` / `npm publish` / tag / push; post-publish smoke | **Only after explicit user approval** in the execution task |
-
-Templates: `.trellis/tasks/templates/release-readiness/` and `.trellis/tasks/templates/release-execution/`. Copy `prd.md`, `design.md`, and `implement.md` into a new task directory; use `handoff-template.md` as the outline for `handoff.md`.
-
-### Release status vocabulary (handoff / verify)
-
-These labels describe **release posture** in `handoff.md` and `verify.md`. They are not `task.json.status` values (contract epoch 1).
-
-| Label | Typical task |
-| --- | --- |
-| **ready to publish** | Readiness — evidence complete; operator may proceed after approval |
-| **not published** | Readiness or aborted execution — no remote release in this task |
-| **published** | Execution — registry and/or git remote actions completed |
-| **blocked** | Required gate failed; do not publish until fixed or waived |
-| **waived** | User explicitly accepted a known blocker (name approver + rationale) |
-| **deferred** | Publish postponed by user (e.g. wait for dogfood / another child) |
-
-Readiness `handoff.md` must include **Ready to publish**, **Not published**, and **Blockers**. Execution `handoff.md` must include **Published** (or **Not published** if aborted) and **Publish approval evidence** in `verify.md`.
-
-### Standard evidence (readiness `verify.md`)
-
-- Version recommendation (current → proposed; channel; npm dist-tag expectation)
-- Changelog / release notes (draft)
-- Migration manifest (id or none)
-- Build evidence, test evidence, pack dry-run evidence
-- Prepublish blockers with disposition: `blocking` / `fixed` / `waived` / `deferred`
-
-Dry-run examples for `@blxzer/cursor-trellis` (from `packages/cli`): `check-manifest-continuity.js`, `release-preflight.js check-versions`, `publish-plan`, `verify-packed-cli`, `npm pack --dry-run`. Do not run `release.js`, `npm publish`, or `git push` during readiness.
-
-### Execution approval gate
-
-Before `release.js`, `npm publish`, `git tag`, `git push`, or GitHub release creation, stop and obtain **explicit user approval**. Record in execution `verify.md` under **Publish approval evidence** (approved version, dist-tag, and which remote actions are allowed). Passing preflight is not approval.
-
-See also `.trellis/spec/Trellis/cli/release-and-vendor.md`.
-
----
-
 ## Customizing Trellis (for forks)
 
 This section is for developers who want to modify the Trellis workflow itself. All customization is done by editing this file; the scripts are parsers only.

package/package.json

@@ -1,14 +1,13 @@
 {
   "name": "@blxzer/cursor-trellis",
-  "version": "0.1.3",
+  "version": "0.2.1",
   "description": "AI capabilities grow like ivy — Trellis provides the structure to guide them along a disciplined path",
   "type": "module",
   "main": "./dist/index.js",
   "types": "./dist/index.d.ts",
   "bin": {
     "trellis": "./bin/trellis.js",
-    "tl": "./bin/trellis.js",
-    "smart-search": "./bin/smart-search.js"
+    "tl": "./bin/trellis.js"
   },
   "publishConfig": {
     "access": "public"
@@ -28,6 +27,7 @@
   "author": "Mindfold LLC",
   "license": "AGPL-3.0-only",
   "dependencies": {
+    "@blxzer/smart-search": "^0.1.0",
     "chalk": "^5.3.0",
     "commander": "^12.1.0",
     "figlet": "^1.9.4",
@@ -35,7 +35,7 @@
     "inquirer": "^9.3.7",
     "undici": "^6.21.0",
     "zod": "^4.4.2",
-    "@blxzer/cursor-trellis-core": "0.1.2"
+    "@blxzer/cursor-trellis-core": "0.2.1"
   },
   "devDependencies": {
     "@eslint/js": "^9.18.0",
@@ -59,43 +59,7 @@
     "scripts/postinstall.js",
     "README.md",
     "CHANGELOG.md",
-    "LICENSE",
-    "vendor/smart-search/LICENSE",
-    "vendor/smart-search/README.md",
-    "vendor/smart-search/README.zh-CN.md",
-    "vendor/smart-search/npm/bin/smart-search.js",
-    "vendor/smart-search/npm/scripts/postinstall.js",
-    "vendor/smart-search/npm/scripts/resolve-prerelease-version.js",
-    "vendor/smart-search/npm/scripts/set-package-version.js",
-    "vendor/smart-search/npm/scripts/sync-python-version.js",
-    "vendor/smart-search/npm/scripts/test-wrapper-repair.js",
-    "vendor/smart-search/npm/scripts/test.js",
-    "vendor/smart-search/package.json",
-    "vendor/smart-search/pyproject.toml",
-    "vendor/smart-search/skills/smart-search-cli/SKILL.md",
-    "vendor/smart-search/skills/smart-search-cli/agents/openai.yaml",
-    "vendor/smart-search/skills/smart-search-cli/examples/batch-search.md",
-    "vendor/smart-search/skills/smart-search-cli/examples/evidence-gathering.md",
-    "vendor/smart-search/skills/smart-search-cli/references/cli-contract.md",
-    "vendor/smart-search/src/smart_search/__init__.py",
-    "vendor/smart-search/src/smart_search/assets/skills/smart-search-cli/SKILL.md",
-    "vendor/smart-search/src/smart_search/assets/skills/smart-search-cli/agents/openai.yaml",
-    "vendor/smart-search/src/smart_search/assets/skills/smart-search-cli/examples/batch-search.md",
-    "vendor/smart-search/src/smart_search/assets/skills/smart-search-cli/examples/evidence-gathering.md",
-    "vendor/smart-search/src/smart_search/assets/skills/smart-search-cli/references/cli-contract.md",
-    "vendor/smart-search/src/smart_search/cli.py",
-    "vendor/smart-search/src/smart_search/config.py",
-    "vendor/smart-search/src/smart_search/logger.py",
-    "vendor/smart-search/src/smart_search/providers/__init__.py",
-    "vendor/smart-search/src/smart_search/providers/base.py",
-    "vendor/smart-search/src/smart_search/providers/context7.py",
-    "vendor/smart-search/src/smart_search/providers/exa.py",
-    "vendor/smart-search/src/smart_search/providers/jina.py",
-    "vendor/smart-search/src/smart_search/providers/openai_compatible.py",
-    "vendor/smart-search/src/smart_search/providers/zhipu.py",
-    "vendor/smart-search/src/smart_search/service.py",
-    "vendor/smart-search/src/smart_search/sources.py",
-    "vendor/smart-search/src/smart_search/utils.py"
+    "LICENSE"
   ],
   "repository": {
     "type": "git",
@@ -106,8 +70,6 @@
     "clean": "node -e \"require('node:fs').rmSync('dist', { recursive: true, force: true })\"",
     "copy-templates": "node scripts/copy-templates.js",
     "postinstall": "node scripts/postinstall.js",
-    "sync:smart-search": "node scripts/sync-smart-search-vendor.js",
-    "check:smart-search": "node scripts/check-smart-search-vendor.js",
     "check:pack-files": "node scripts/check-cli-pack-files.js",
     "check:release-pack": "node scripts/check-release-pack-contents.js",
     "check:router-copy-sync": "node scripts/check-router-copy-sync.js",
@@ -126,7 +88,6 @@
     "typecheck": "tsc --noEmit",
     "lint:py": "basedpyright",
     "lint:all": "pnpm lint && pnpm lint:py",
-    "prepublishOnly_disabled": "pnpm test && pnpm run build && pnpm run copy:release-assets",
     "release": "node scripts/release.js patch",
     "release:minor": "node scripts/release.js minor",
     "release:major": "node scripts/release.js major",

package/bin/smart-search.js

@@ -1,79 +0,0 @@
-#!/usr/bin/env node
-
-import { spawn, spawnSync } from "node:child_process";
-import fs from "node:fs";
-import path from "node:path";
-import { fileURLToPath } from "node:url";
-
-const packageRoot = path.resolve(
-  path.dirname(fileURLToPath(import.meta.url)),
-  "..",
-);
-const vendorRoot = path.join(packageRoot, "vendor", "smart-search");
-const callerCwd = process.env.INIT_CWD || process.cwd();
-const venvDir = path.join(vendorRoot, ".smart-search-python");
-const pythonPath =
-  process.platform === "win32"
-    ? path.join(venvDir, "Scripts", "python.exe")
-    : path.join(venvDir, "bin", "python");
-const postinstall = path.join(vendorRoot, "npm", "scripts", "postinstall.js");
-
-function fail(message, code = 5) {
-  console.error(message);
-  process.exit(code);
-}
-
-if (!fs.existsSync(path.join(vendorRoot, "pyproject.toml"))) {
-  fail(
-    "Trellis smart-search runtime is missing. Reinstall @blxzer/cursor-trellis or run the package build again.",
-  );
-}
-
-if (!fs.existsSync(pythonPath)) {
-  if (!fs.existsSync(postinstall)) {
-    fail(
-      `smart-search Python runtime is missing and repair script was not found: ${postinstall}`,
-    );
-  }
-
-  console.error("smart-search Python runtime is missing; attempting repair...");
-  const repaired = spawnSync(process.execPath, [postinstall], {
-    cwd: vendorRoot,
-    stdio: "inherit",
-    windowsHide: true,
-  });
-  if (repaired.error) {
-    fail(`smart-search runtime repair failed: ${repaired.error.message}`);
-  }
-  if (repaired.status !== 0 || !fs.existsSync(pythonPath)) {
-    console.error("Trellis could not find the smart-search Python runtime.");
-    console.error(`Expected: ${pythonPath}`);
-    console.error("Repair it by reinstalling Trellis or running:");
-    console.error("  node scripts/postinstall.js");
-    process.exit(repaired.status || 5);
-  }
-}
-
-const child = spawn(pythonPath, ["-m", "smart_search.cli", ...process.argv.slice(2)], {
-  cwd: callerCwd,
-  stdio: "inherit",
-  env: {
-    ...process.env,
-    SMART_SEARCH_PACKAGE_ROOT: vendorRoot,
-    PYTHONIOENCODING: process.env.PYTHONIOENCODING || "utf-8",
-    PYTHONUTF8: process.env.PYTHONUTF8 || "1",
-  },
-  windowsHide: true,
-});
-
-child.on("error", (error) => {
-  fail(`Failed to start smart-search: ${error.message}`);
-});
-
-child.on("close", (code, signal) => {
-  if (signal) {
-    process.kill(process.pid, signal);
-    return;
-  }
-  process.exit(code ?? 5);
-});

package/dist/templates/trellis/scripts/aggregate_retrieval_telemetry.py

@@ -1,342 +0,0 @@
-#!/usr/bin/env python3
-"""Aggregate per-query retrieval telemetry JSONL into summary metrics (schema v2)."""
-
-from __future__ import annotations
-
-import argparse
-import json
-import sys
-from pathlib import Path
-from typing import Any
-
-_SCRIPT_DIR = Path(__file__).resolve().parent
-if str(_SCRIPT_DIR) not in sys.path:
-    sys.path.insert(0, str(_SCRIPT_DIR))
-
-from common.retrieval_tool_classification import (  # noqa: E402
-    classify_tool_calls,
-    semantic_routes_in_plan,
-    structural_routes_in_plan,
-)
-
-SCHEMA_VERSION = 2
-
-SEMANTIC_OUTCOMES = (
-    "success",
-    "partial",
-    "resource_exhausted",
-    "timeout",
-    "not_configured",
-    "unavailable",
-    "not_run",
-    "unknown",
-)
-
-SEMANTIC_SKIP_REASONS = (
-    "rg_corrob_sufficient",
-    "rg_empty_semantic_required",
-    "trap_only_semantic_required",
-    "not_in_plan",
-    "not_applicable",
-    "adapter_unavailable",
-    "agent_stopped_early",
-    "unknown",
-    "manual_not_recorded",
-)
-
-
-def _rate(count: int, total: int) -> float:
-    return 0.0 if total == 0 else count / total
-
-
-def _avg(values: list[float]) -> float:
-    return 0.0 if not values else sum(values) / len(values)
-
-
-def compute_compliance_score(record: dict[str, Any]) -> float:
-    earned = 0.0
-    possible = 0.0
-
-    if record.get("structural_in_plan"):
-        possible += 1.0
-        if record.get("codegraph_executed"):
-            earned += 1.0
-
-    if record.get("semantic_in_plan"):
-        possible += 1.0
-        if record.get("semantic_executed") or record.get("semantic_skip_reason") == "rg_corrob_sufficient":
-            earned += 1.0
-
-    possible += 1.0
-    if record.get("read_verification_done"):
-        earned += 1.0
-
-    if record.get("plan_block_in_prompt"):
-        possible += 0.25
-        if record.get("router_cli_invoked"):
-            earned += 0.25
-
-    return 1.0 if possible == 0 else earned / possible
-
-
-def migrate_record(raw: dict[str, Any]) -> dict[str, Any]:
-    routes = raw.get("routes") or raw.get("routes_in_plan") or []
-    if not isinstance(routes, list):
-        routes = []
-    tools = raw.get("tools_called") or []
-    if not isinstance(tools, list):
-        tools = []
-    platform = str(raw.get("platform", "cursor"))
-    classified = classify_tool_calls([str(t) for t in tools], platform=platform)
-
-    structural_in = raw.get("structural_in_plan")
-    if not isinstance(structural_in, bool):
-        structural_in = structural_routes_in_plan([str(r) for r in routes])
-
-    semantic_in = raw.get("semantic_in_plan")
-    if not isinstance(semantic_in, bool):
-        semantic_in = semantic_routes_in_plan([str(r) for r in routes])
-
-    record: dict[str, Any] = {
-        "schema_version": SCHEMA_VERSION,
-        "query_id": str(raw.get("query_id", "")),
-        "dataset": str(raw.get("dataset", "unknown")),
-        "query_text": str(raw.get("query_text", "")),
-        "run_id": str(raw.get("run_id", "")),
-        "platform": platform,
-        "semantic_in_plan": bool(semantic_in),
-        "semantic_order": raw.get("semantic_order")
-        if isinstance(raw.get("semantic_order"), int)
-        else None,
-        "structural_in_plan": bool(structural_in),
-        "codegraph_in_plan": bool(raw.get("codegraph_in_plan", structural_in)),
-        "fallback_hint_present": bool(raw.get("fallback_hint_present")),
-        "intents": list(raw.get("intents") or []),
-        "routes": [str(r) for r in routes],
-        "routes_in_plan": [str(r) for r in routes],
-        "project_file_count": raw.get("project_file_count")
-        if isinstance(raw.get("project_file_count"), int)
-        else None,
-        "tools_called": classified.tools_called,
-        "grep_count": raw.get("grep_count")
-        if isinstance(raw.get("grep_count"), int)
-        else classified.grep_count,
-        "read_count": raw.get("read_count")
-        if isinstance(raw.get("read_count"), int)
-        else classified.read_count,
-        "codegraph_attempted": bool(
-            raw.get("codegraph_attempted", classified.codegraph_attempted)
-        ),
-        "codegraph_executed": bool(
-            raw.get("codegraph_executed", classified.codegraph_executed)
-        ),
-        "router_cli_invoked": bool(
-            raw.get("router_cli_invoked", classified.router_cli_invoked)
-        ),
-        "plan_block_in_prompt": bool(raw.get("plan_block_in_prompt")),
-        "read_verification_done": bool(
-            raw.get("read_verification_done", classified.read_count > 0)
-        ),
-        "semantic_attempted": bool(
-            raw.get("semantic_attempted", classified.semantic_attempted)
-        ),
-        "semantic_executed": bool(
-            raw.get("semantic_executed", classified.semantic_executed)
-        ),
-        "semantic_outcome": str(raw.get("semantic_outcome", "unknown")),
-        "semantic_success": bool(raw.get("semantic_success")),
-        "semantic_skip_reason": raw.get("semantic_skip_reason"),
-        "rg_candidate_count": raw.get("rg_candidate_count")
-        if isinstance(raw.get("rg_candidate_count"), int)
-        else None,
-        "rg_corrob_status": str(raw.get("rg_corrob_status", "unknown")),
-        "trap_only": bool(raw.get("trap_only")),
-        "corroborated_files": list(raw.get("corroborated_files") or []),
-        "adapter_errors": list(raw.get("adapter_errors") or []),
-        "candidate_pool_recall": raw.get("candidate_pool_recall")
-        if isinstance(raw.get("candidate_pool_recall"), (int, float))
-        else None,
-        "final_top_k_recall": raw.get("final_top_k_recall")
-        if isinstance(raw.get("final_top_k_recall"), (int, float))
-        else None,
-        "answer_score": raw.get("answer_score")
-        if isinstance(raw.get("answer_score"), (int, float))
-        else None,
-        "compliance_score": raw.get("compliance_score")
-        if isinstance(raw.get("compliance_score"), (int, float))
-        else None,
-        "platform_semantic_executed": bool(
-            raw.get("platform_semantic_executed", classified.platform_semantic_executed)
-        ),
-        "fast_context_count": int(raw.get("fast_context_count", classified.fast_context_count)),
-        "cursor_fast_context_misuse": bool(
-            raw.get("cursor_fast_context_misuse", classified.cursor_fast_context_misuse)
-        ),
-    }
-    if record["compliance_score"] is None:
-        record["compliance_score"] = compute_compliance_score(record)
-    return record
-
-
-def derive_metrics(records: list[dict[str, Any]]) -> dict[str, Any]:
-    total = len(records)
-    outcome_counts = {k: 0 for k in SEMANTIC_OUTCOMES}
-    skip_counts = {k: 0 for k in SEMANTIC_SKIP_REASONS}
-
-    semantic_plan = semantic_exec = semantic_attempt = semantic_success = 0
-    platform_sem_exec = fast_context_total = fast_context_misuse = 0
-    codegraph_plan = codegraph_exec = router_cli = plan_block = read_verify = 0
-    compliance_scores: list[float] = []
-    answer_scores: list[float] = []
-    pool_recalls: list[float] = []
-    topk_recalls: list[float] = []
-
-    for rec in records:
-        if rec.get("semantic_in_plan"):
-            semantic_plan += 1
-        if rec.get("semantic_executed"):
-            semantic_exec += 1
-        if rec.get("platform_semantic_executed"):
-            platform_sem_exec += 1
-        fc = rec.get("fast_context_count")
-        if isinstance(fc, int) and fc > 0:
-            fast_context_total += fc
-        if rec.get("cursor_fast_context_misuse"):
-            fast_context_misuse += 1
-        if rec.get("semantic_attempted"):
-            semantic_attempt += 1
-        if rec.get("semantic_success") and rec.get("semantic_outcome") == "success":
-            semantic_success += 1
-        if rec.get("codegraph_in_plan"):
-            codegraph_plan += 1
-        if rec.get("codegraph_executed"):
-            codegraph_exec += 1
-        if rec.get("router_cli_invoked"):
-            router_cli += 1
-        if rec.get("plan_block_in_prompt"):
-            plan_block += 1
-        if rec.get("read_verification_done"):
-            read_verify += 1
-
-        outcome = str(rec.get("semantic_outcome", "unknown"))
-        if outcome in outcome_counts:
-            outcome_counts[outcome] += 1
-        skip = rec.get("semantic_skip_reason")
-        if skip and skip in skip_counts:
-            skip_counts[skip] += 1
-
-        cs = rec.get("compliance_score")
-        if isinstance(cs, (int, float)):
-            compliance_scores.append(float(cs))
-        ans = rec.get("answer_score")
-        if isinstance(ans, (int, float)):
-            answer_scores.append(float(ans))
-        cpr = rec.get("candidate_pool_recall")
-        if isinstance(cpr, (int, float)):
-            pool_recalls.append(float(cpr))
-        ftr = rec.get("final_top_k_recall")
-        if isinstance(ftr, (int, float)):
-            topk_recalls.append(float(ftr))
-
-    avg_pool = _avg(pool_recalls)
-    avg_topk = _avg(topk_recalls)
-    recall_drop = 0.0 if avg_pool <= 0 else 1.0 - avg_topk / avg_pool
-
-    return {
-        "schema_version": SCHEMA_VERSION,
-        "total_queries": total,
-        "semantic_plan_count": semantic_plan,
-        "semantic_exec_count": semantic_exec,
-        "semantic_attempt_count": semantic_attempt,
-        "semantic_exec_success_count": semantic_success,
-        "semantic_plan_rate": _rate(semantic_plan, total),
-        "semantic_exec_rate": _rate(semantic_exec, total),
-        "platform_semantic_exec_count": platform_sem_exec,
-        "platform_semantic_exec_rate": _rate(platform_sem_exec, total),
-        "fast_context_invocation_total": fast_context_total,
-        "cursor_fast_context_misuse_count": fast_context_misuse,
-        "cursor_fast_context_misuse_rate": _rate(fast_context_misuse, total),
-        "semantic_attempt_rate": _rate(semantic_attempt, total),
-        "semantic_exec_success_rate": _rate(semantic_success, total),
-        "codegraph_plan_count": codegraph_plan,
-        "codegraph_exec_count": codegraph_exec,
-        "codegraph_plan_rate": _rate(codegraph_plan, total),
-        "codegraph_exec_rate": _rate(codegraph_exec, total),
-        "router_cli_count": router_cli,
-        "router_cli_rate": _rate(router_cli, total),
-        "plan_block_count": plan_block,
-        "plan_block_rate": _rate(plan_block, total),
-        "read_verification_count": read_verify,
-        "read_verification_rate": _rate(read_verify, total),
-        "avg_compliance_score": _avg(compliance_scores),
-        "avg_answer_score": _avg(answer_scores),
-        "semantic_outcome_counts": outcome_counts,
-        "semantic_skip_reason_counts": skip_counts,
-        "avg_candidate_pool_recall": avg_pool,
-        "avg_final_top_k_recall": avg_topk,
-        "recall_drop_rate": recall_drop,
-    }
-
-
-def load_jsonl(path: Path) -> list[dict[str, Any]]:
-    records: list[dict[str, Any]] = []
-    for line_no, line in enumerate(path.read_text(encoding="utf-8").splitlines(), start=1):
-        stripped = line.strip()
-        if not stripped or stripped.startswith("#"):
-            continue
-        try:
-            raw = json.loads(stripped)
-        except json.JSONDecodeError as exc:
-            raise ValueError(f"{path}:{line_no}: invalid JSON: {exc}") from exc
-        if not isinstance(raw, dict):
-            raise ValueError(f"{path}:{line_no}: expected JSON object per line")
-        records.append(migrate_record(raw))
-    return records
-
-
-def main() -> int:
-    parser = argparse.ArgumentParser(description=__doc__)
-    parser.add_argument(
-        "jsonl",
-        type=Path,
-        help="Per-query telemetry JSONL (one object per line)",
-    )
-    parser.add_argument(
-        "--markdown",
-        action="store_true",
-        help="Print a short markdown summary for eval reports",
-    )
-    args = parser.parse_args()
-
-    if not args.jsonl.is_file():
-        print(f"error: file not found: {args.jsonl}", file=sys.stderr)
-        return 1
-
-    records = load_jsonl(args.jsonl)
-    metrics = derive_metrics(records)
-    payload = {"metrics": metrics, "record_count": len(records)}
-
-    if args.markdown:
-        m = metrics
-        print("## Retrieval telemetry (derived from JSONL)\n")
-        print("| Metric | Value |")
-        print("| --- | ---: |")
-        print(f"| total_queries | {m['total_queries']} |")
-        print(f"| semantic_plan_rate | {m['semantic_plan_rate']:.1%} |")
-        print(f"| semantic_exec_rate | {m['semantic_exec_rate']:.1%} |")
-        print(f"| platform_semantic_exec_rate | {m.get('platform_semantic_exec_rate', 0):.1%} |")
-        print(f"| cursor_fast_context_misuse_rate | {m.get('cursor_fast_context_misuse_rate', 0):.1%} |")
-        print(f"| codegraph_plan_rate | {m['codegraph_plan_rate']:.1%} |")
-        print(f"| codegraph_exec_rate | {m['codegraph_exec_rate']:.1%} |")
-        print(f"| router_cli_rate | {m['router_cli_rate']:.1%} |")
-        print(f"| avg_compliance_score | {m['avg_compliance_score']:.3f} |")
-        print(f"| avg_answer_score | {m['avg_answer_score']:.3f} |")
-        print("\nDo not hand-copy plan/exec rates; regenerate from JSONL with this script.")
-    else:
-        print(json.dumps(payload, indent=2, ensure_ascii=False))
-
-    return 0
-
-
-if __name__ == "__main__":
-    raise SystemExit(main())