feat: expand family suite and publish eval dashboard

Makefile

@@ -1,6 +1,6 @@
 PYTHON ?= python3
 
-.PHONY: eval eval-suite failure-regression-check package-check package-failure-check snapshot-check validate lint test clean
+.PHONY: eval eval-suite results-panel failure-regression-check package-check package-failure-check snapshot-check validate lint test clean
 
 eval:
 	$(PYTHON) scripts/trigger_eval.py --description-file evals/improved_description.txt --cases evals/trigger_cases.json --baseline-description-file evals/baseline_description.txt
@@ -8,6 +8,9 @@ eval:
 eval-suite:
 	$(PYTHON) scripts/run_eval_suite.py
 
+results-panel:
+	$(PYTHON) scripts/render_eval_dashboard.py
+
 failure-regression-check:
 	$(PYTHON) tests/verify_failure_regressions.py
 

README.md

@@ -46,9 +46,46 @@ make test
 
 ## Results
 
-- `make test` passes locally
-- semantic trigger eval: `0` false positives, `0` false negatives on the current regression set
-- eval suite: train / dev / holdout all pass, with family-based reporting
+The homepage panel below is generated from the current eval suite so the family-level outcome is visible without opening raw JSON.
+
+<!-- BEGIN:EVAL_RESULTS -->
+- regression corpus: `66` prompts across `21` families
+- aggregate result: `0` false positives, `0` false negatives, average precision `1.0`, average recall `1.0`
+- suite status:
+
+| Suite | Cases | FP | FN | Precision | Recall |
+| --- | ---: | ---: | ---: | ---: | ---: |
+| train | 31 | 0 | 0 | 1.0 | 1.0 |
+| dev | 22 | 0 | 0 | 1.0 | 1.0 |
+| holdout | 13 | 0 | 0 | 1.0 | 1.0 |
+
+| Family | Cases | Pass Rate |
+| --- | ---: | ---: |
+| `brainstorm_only` | 2 | 1.0 |
+| `brainstorm_vs_build` | 1 | 1.0 |
+| `complex_multi_asset` | 3 | 1.0 |
+| `document_export_vs_agent_skill` | 4 | 1.0 |
+| `document_only` | 3 | 1.0 |
+| `explain_not_package` | 1 | 1.0 |
+| `explain_only` | 5 | 1.0 |
+| `future_outline_vs_build` | 4 | 1.0 |
+| `iterate_existing_skill` | 5 | 1.0 |
+| `long_context_document_only` | 3 | 1.0 |
+| `long_context_near_neighbor` | 3 | 1.0 |
+| `long_context_summary_only` | 2 | 1.0 |
+| `long_context_trigger` | 4 | 1.0 |
+| `meta_skill_creation` | 1 | 1.0 |
+| `one_off_vs_reusable` | 2 | 1.0 |
+| `package_for_team` | 2 | 1.0 |
+| `paraphrase_trigger` | 5 | 1.0 |
+| `partial_scaffold_not_full_skill` | 4 | 1.0 |
+| `summary_only` | 3 | 1.0 |
+| `translate_only` | 4 | 1.0 |
+| `workflow_to_skill` | 5 | 1.0 |
+
+Full reports: [reports/eval_suite.json](reports/eval_suite.json) and [reports/family_summary.md](reports/family_summary.md)
+<!-- END:EVAL_RESULTS -->
+
 - packaging validation: `openai`, `claude`, and `generic` targets pass contract checks
 - packaging failure fixtures: invalid metadata, invalid YAML, and unsupported targets fail as expected
 - failure library regressions: anti-pattern families pass automated checks

evals/README.md

@@ -12,6 +12,7 @@ Contents:
 - `sample_trigger_report.json`: example comparison output using the current recommended threshold
 - `failure-cases.md`: current weak spots and regression targets
 - `packaging_expectations.json`: required packaging behaviors for supported targets
+- `../reports/`: generated suite JSON plus the homepage-visible family summary panel source
 
 Use:
 
@@ -19,6 +20,7 @@ Use:
 python3 scripts/trigger_eval.py --description-file evals/improved_description.txt --cases evals/trigger_cases.json
 python3 scripts/trigger_eval.py --description-file evals/improved_description.txt --cases evals/trigger_cases.json --baseline-description-file evals/baseline_description.txt
 python3 scripts/run_eval_suite.py
+python3 scripts/render_eval_dashboard.py
 python3 tests/verify_failure_regressions.py
 python3 scripts/cross_packager.py . --platform openai --platform claude --expectations evals/packaging_expectations.json --zip
 python3 tests/verify_packager_failures.py
@@ -33,5 +35,7 @@ Regression scope now includes:
 - mixed-intent negatives
 - explicit "do not build a skill" negatives
 - semantic exclusion cases such as one-off, document-only, and future-outline prompts
+- paraphrase families that avoid the original wording while preserving the same trigger intent
+- long-context contamination cases where build intent or no-build intent appears after unrelated setup text
 - family-based reporting across workflow-to-skill, iterate-existing-skill, document-only, one-off, and future-outline cases
 - holdout verification

evals/dev/trigger_cases.json

@@ -29,6 +29,18 @@
     {
       "text": "Tighten the trigger boundary on this existing skill and prepare it for team distribution.",
       "family": "iterate_existing_skill"
+    },
+    {
+      "text": "Turn the SOP, transcript, and cleanup notes into a maintained skill for the team library.",
+      "family": "paraphrase_trigger"
+    },
+    {
+      "text": "Codify this workflow into a shared capability package and validate the route boundary before distribution.",
+      "family": "paraphrase_trigger"
+    },
+    {
+      "text": "After the long handoff notes below, build a skill package for team reuse. The materials include a transcript, rough notes, and an onboarding flow, and the result should include references, scripts, and validation.",
+      "family": "long_context_trigger"
     }
   ],
   "should_not_trigger": [
@@ -51,6 +63,14 @@
     {
       "text": "Summarize this workflow and list the main points only.",
       "family": "summary_only"
+    },
+    {
+      "text": "This is a long incident write-up. Keep it as documentation only, explain it in simpler language, and do not convert it into a reusable capability.",
+      "family": "long_context_document_only"
+    },
+    {
+      "text": "I pasted a very long process note. For now I only need a recap and a short summary, not a skill package.",
+      "family": "long_context_summary_only"
     }
   ],
   "near_neighbor": [
@@ -73,6 +93,14 @@
     {
       "text": "Polish these notes into a reusable-looking document without turning them into an agent skill.",
       "family": "document_export_vs_agent_skill"
+    },
+    {
+      "text": "Turn the workflow below into a wiki-ready reference document and template only. We may build a skill later, but not now.",
+      "family": "long_context_near_neighbor"
+    },
+    {
+      "text": "Shape this idea for later and produce a future-ready outline, but do not build the skill package yet.",
+      "family": "future_outline_vs_build"
     }
   ]
 }

evals/holdout/trigger_cases.json

@@ -21,6 +21,14 @@
     {
       "text": "Turn this recurring onboarding process into a reusable skill library entry with validation notes.",
       "family": "workflow_to_skill"
+    },
+    {
+      "text": "Codify this quarterly release routine into a reusable skill package with scripts, references, and packaging checks.",
+      "family": "paraphrase_trigger"
+    },
+    {
+      "text": "Below is a noisy transcript plus release notes and prompt history. Ignore the chatter and turn this playbook into an agent capability package for the team library.",
+      "family": "long_context_trigger"
     }
   ],
   "should_not_trigger": [
@@ -31,6 +39,14 @@
     {
       "text": "Explain this release checklist in simpler language; do not package it.",
       "family": "explain_only"
+    },
+    {
+      "text": "Keep this release workflow as documentation only for the wiki and produce a reference document, not for agent execution.",
+      "family": "long_context_document_only"
+    },
+    {
+      "text": "Translate this long runbook into Russian and keep it as documentation only with no skill package.",
+      "family": "translate_only"
     }
   ],
   "near_neighbor": [
@@ -45,6 +61,10 @@
     {
       "text": "Turn this into a document package for the wiki, not an agent skill.",
       "family": "document_export_vs_agent_skill"
+    },
+    {
+      "text": "From the long workflow below, create a template only and a checklist for discussion only. No skill package.",
+      "family": "long_context_near_neighbor"
     }
   ]
 }

evals/semantic_config.json

@@ -19,7 +19,13 @@
         "meta skill",
         "meta-skill",
         "turn this into a skill",
-        "turn all of that into one reusable skill package"
+        "turn all of that into one reusable skill package",
+        "codify this workflow",
+        "codify this routine",
+        "operationalize this workflow as a reusable skill",
+        "formalize this process into a reusable capability",
+        "turn this playbook into an agent capability package",
+        "convert this into a maintained skill"
       ]
     },
     "transform_workflow": {
@@ -33,7 +39,15 @@
         "operations checklist",
         "internal process note",
         "existing skill draft",
-        "rough notes"
+        "rough notes",
+        "standard operating procedure",
+        "sop",
+        "transcript",
+        "workflow fragments",
+        "release notes",
+        "onboarding flow",
+        "support escalation routine",
+        "quarterly release routine"
       ]
     },
     "reuse_package": {
@@ -47,7 +61,13 @@
         "share it with the team",
         "team reuse",
         "internal workflows",
-        "production-ready"
+        "production-ready",
+        "library entry",
+        "shared capability",
+        "team library",
+        "ops library",
+        "maintained",
+        "operationalize"
       ]
     },
     "eval_optimize": {
@@ -60,7 +80,12 @@
         "benchmark",
         "validation",
         "validate",
-        "packaging checks"
+        "packaging checks",
+        "false positives",
+        "false negatives",
+        "stress test",
+        "route boundary",
+        "hardening"
       ]
     },
     "iterate_existing_skill": {
@@ -73,7 +98,11 @@
         "optimize skill triggering",
         "package this skill",
         "sharing it with the team",
-        "before sharing it with the team"
+        "before sharing it with the team",
+        "tighten the boundary",
+        "revise this skill",
+        "draft skill",
+        "harden this skill for library reuse"
       ]
     },
     "multi_asset": {
@@ -172,7 +201,12 @@
         "not a skill yet",
         "plain note",
         "keep it as a plain note",
-        "no packaging needed"
+        "no packaging needed",
+        "not for agent execution",
+        "no skill package",
+        "do not convert it into a reusable capability",
+        "keep it as documentation only",
+        "discussion only"
       ]
     },
     "document_only": {
@@ -187,7 +221,10 @@
         "article",
         "blog title",
         "headline",
-        "landing page"
+        "landing page",
+        "reference document",
+        "knowledge base",
+        "template only"
       ]
     },
     "future_outline": {
@@ -197,7 +234,10 @@
         "future skill",
         "possible future skill",
         "before we decide whether to build",
-        "outline for a possible future skill"
+        "outline for a possible future skill",
+        "maybe later",
+        "for later",
+        "future-ready outline"
       ]
     }
   }

evals/train/trigger_cases.json

@@ -41,6 +41,22 @@
     {
       "text": "Standardize this recurring workflow as a shareable skill package with references.",
       "family": "package_for_team"
+    },
+    {
+      "text": "Codify this recurring support escalation workflow into a maintained skill for the ops library.",
+      "family": "paraphrase_trigger"
+    },
+    {
+      "text": "Operationalize this workflow as a reusable skill with validation notes and team library metadata.",
+      "family": "paraphrase_trigger"
+    },
+    {
+      "text": "Below are release notes, rough notes, and a transcript from the last handoff. Ignore the noise and turn this runbook into a reusable agent skill with references and scripts so the team can reuse it every quarter.",
+      "family": "long_context_trigger"
+    },
+    {
+      "text": "We have a standard operating procedure, checklist, and prompt history scattered across docs. Formalize this process into a reusable capability, add validation, and package it for team reuse.",
+      "family": "long_context_trigger"
     }
   ],
   "should_not_trigger": [
@@ -75,6 +91,18 @@
     {
       "text": "Explain the difference between a runbook and a workflow.",
       "family": "explain_only"
+    },
+    {
+      "text": "Below is a long SOP, but I only want a summary for leadership and a recap of the main points. Do not build anything reusable.",
+      "family": "long_context_summary_only"
+    },
+    {
+      "text": "Keep this as documentation only for the knowledge base and rewrite it as a reference document.",
+      "family": "long_context_document_only"
+    },
+    {
+      "text": "Translate this long onboarding flow into French and keep it as documentation only with no skill package.",
+      "family": "translate_only"
     }
   ],
   "near_neighbor": [
@@ -101,6 +129,14 @@
     {
       "text": "Create a reusable-looking checklist, but keep it as a plain note instead of a skill.",
       "family": "partial_scaffold_not_full_skill"
+    },
+    {
+      "text": "We may convert this workflow later, but for now keep it as documentation only and produce a reference document.",
+      "family": "long_context_near_neighbor"
+    },
+    {
+      "text": "Draft a template only from this process and stop short of making a full skill package.",
+      "family": "partial_scaffold_not_full_skill"
     }
   ]
 }

evals/trigger_cases.json

@@ -73,6 +73,42 @@
     {
       "text": "Turn this recurring onboarding process into a reusable skill library entry with validation notes.",
       "family": "workflow_to_skill"
+    },
+    {
+      "text": "Codify this recurring support escalation workflow into a maintained skill for the ops library.",
+      "family": "paraphrase_trigger"
+    },
+    {
+      "text": "Operationalize this workflow as a reusable skill with validation notes and team library metadata.",
+      "family": "paraphrase_trigger"
+    },
+    {
+      "text": "Turn the SOP, transcript, and cleanup notes into a maintained skill for the team library.",
+      "family": "paraphrase_trigger"
+    },
+    {
+      "text": "Codify this workflow into a shared capability package and validate the route boundary before distribution.",
+      "family": "paraphrase_trigger"
+    },
+    {
+      "text": "Codify this quarterly release routine into a reusable skill package with scripts, references, and packaging checks.",
+      "family": "paraphrase_trigger"
+    },
+    {
+      "text": "Below are release notes, rough notes, and a transcript from the last handoff. Ignore the noise and turn this runbook into a reusable agent skill with references and scripts so the team can reuse it every quarter.",
+      "family": "long_context_trigger"
+    },
+    {
+      "text": "We have a standard operating procedure, checklist, and prompt history scattered across docs. Formalize this process into a reusable capability, add validation, and package it for team reuse.",
+      "family": "long_context_trigger"
+    },
+    {
+      "text": "After the long handoff notes below, build a skill package for team reuse. The materials include a transcript, rough notes, and an onboarding flow, and the result should include references, scripts, and validation.",
+      "family": "long_context_trigger"
+    },
+    {
+      "text": "Below is a noisy transcript plus release notes and prompt history. Ignore the chatter and turn this playbook into an agent capability package for the team library.",
+      "family": "long_context_trigger"
     }
   ],
   "should_not_trigger": [
@@ -135,6 +171,34 @@
     {
       "text": "Explain this release checklist in simpler language; do not package it.",
       "family": "explain_only"
+    },
+    {
+      "text": "Below is a long SOP, but I only want a summary for leadership and a recap of the main points. Do not build anything reusable.",
+      "family": "long_context_summary_only"
+    },
+    {
+      "text": "Keep this as documentation only for the knowledge base and rewrite it as a reference document.",
+      "family": "long_context_document_only"
+    },
+    {
+      "text": "Translate this long onboarding flow into French and keep it as documentation only with no skill package.",
+      "family": "translate_only"
+    },
+    {
+      "text": "This is a long incident write-up. Keep it as documentation only, explain it in simpler language, and do not convert it into a reusable capability.",
+      "family": "long_context_document_only"
+    },
+    {
+      "text": "I pasted a very long process note. For now I only need a recap and a short summary, not a skill package.",
+      "family": "long_context_summary_only"
+    },
+    {
+      "text": "Keep this release workflow as documentation only for the wiki and produce a reference document, not for agent execution.",
+      "family": "long_context_document_only"
+    },
+    {
+      "text": "Translate this long runbook into Russian and keep it as documentation only with no skill package.",
+      "family": "translate_only"
     }
   ],
   "near_neighbor": [
@@ -193,6 +257,26 @@
     {
       "text": "Turn this into a document package for the wiki, not an agent skill.",
       "family": "document_export_vs_agent_skill"
+    },
+    {
+      "text": "We may convert this workflow later, but for now keep it as documentation only and produce a reference document.",
+      "family": "long_context_near_neighbor"
+    },
+    {
+      "text": "Draft a template only from this process and stop short of making a full skill package.",
+      "family": "partial_scaffold_not_full_skill"
+    },
+    {
+      "text": "Turn the workflow below into a wiki-ready reference document and template only. We may build a skill later, but not now.",
+      "family": "long_context_near_neighbor"
+    },
+    {
+      "text": "Shape this idea for later and produce a future-ready outline, but do not build the skill package yet.",
+      "family": "future_outline_vs_build"
+    },
+    {
+      "text": "From the long workflow below, create a template only and a checklist for discussion only. No skill package.",
+      "family": "long_context_near_neighbor"
     }
   ]
 }

reports/family_summary.md

@@ -0,0 +1,34 @@
+# Eval Family Summary
+
+- total cases: `66`
+- families: `21`
+- false positives: `0`
+- false negatives: `0`
+- average precision: `1.0`
+- average recall: `1.0`
+
+## Family Results
+
+| Family | Cases | Passed | FP | FN | Pass Rate |
+| --- | ---: | ---: | ---: | ---: | ---: |
+| `brainstorm_only` | 2 | 2 | 0 | 0 | 1.0 |
+| `brainstorm_vs_build` | 1 | 1 | 0 | 0 | 1.0 |
+| `complex_multi_asset` | 3 | 3 | 0 | 0 | 1.0 |
+| `document_export_vs_agent_skill` | 4 | 4 | 0 | 0 | 1.0 |
+| `document_only` | 3 | 3 | 0 | 0 | 1.0 |
+| `explain_not_package` | 1 | 1 | 0 | 0 | 1.0 |
+| `explain_only` | 5 | 5 | 0 | 0 | 1.0 |
+| `future_outline_vs_build` | 4 | 4 | 0 | 0 | 1.0 |
+| `iterate_existing_skill` | 5 | 5 | 0 | 0 | 1.0 |
+| `long_context_document_only` | 3 | 3 | 0 | 0 | 1.0 |
+| `long_context_near_neighbor` | 3 | 3 | 0 | 0 | 1.0 |
+| `long_context_summary_only` | 2 | 2 | 0 | 0 | 1.0 |
+| `long_context_trigger` | 4 | 4 | 0 | 0 | 1.0 |
+| `meta_skill_creation` | 1 | 1 | 0 | 0 | 1.0 |
+| `one_off_vs_reusable` | 2 | 2 | 0 | 0 | 1.0 |
+| `package_for_team` | 2 | 2 | 0 | 0 | 1.0 |
+| `paraphrase_trigger` | 5 | 5 | 0 | 0 | 1.0 |
+| `partial_scaffold_not_full_skill` | 4 | 4 | 0 | 0 | 1.0 |
+| `summary_only` | 3 | 3 | 0 | 0 | 1.0 |
+| `translate_only` | 4 | 4 | 0 | 0 | 1.0 |
+| `workflow_to_skill` | 5 | 5 | 0 | 0 | 1.0 |

scripts/render_eval_dashboard.py

@@ -0,0 +1,126 @@
+#!/usr/bin/env python3
+import argparse
+import json
+import subprocess
+import sys
+from pathlib import Path
+
+
+BEGIN_MARKER = "<!-- BEGIN:EVAL_RESULTS -->"
+END_MARKER = "<!-- END:EVAL_RESULTS -->"
+
+
+def run_eval_suite(args: argparse.Namespace) -> dict:
+    proc = subprocess.run(
+        [
+            sys.executable,
+            "scripts/run_eval_suite.py",
+            "--eval-dir",
+            args.eval_dir,
+            "--description-file",
+            args.description_file,
+            "--baseline-description-file",
+            args.baseline_description_file,
+        ],
+        capture_output=True,
+        text=True,
+        check=True,
+    )
+    return json.loads(proc.stdout)
+
+
+def render_readme_panel(report: dict) -> str:
+    summary = report["summary"]
+    suites = report["suites"]
+    families = report["family_summary"]
+
+    lines = [
+        BEGIN_MARKER,
+        f"- regression corpus: `{summary['total_cases']}` prompts across `{summary['family_count']}` families",
+        f"- aggregate result: `{summary['false_positives']}` false positives, `{summary['false_negatives']}` false negatives, average precision `{summary['average_precision']}`, average recall `{summary['average_recall']}`",
+        "- suite status:",
+        "",
+        "| Suite | Cases | FP | FN | Precision | Recall |",
+        "| --- | ---: | ---: | ---: | ---: | ---: |",
+    ]
+    for suite_name in ("train", "dev", "holdout"):
+        suite = suites[suite_name]
+        cases = sum(bucket["total"] for bucket in suite["bucket_stats"].values())
+        lines.append(
+            f"| {suite_name} | {cases} | {suite['false_positives']} | {suite['false_negatives']} | {suite['precision']} | {suite['recall']} |"
+        )
+
+    lines.extend(
+        [
+            "",
+            "| Family | Cases | Pass Rate |",
+            "| --- | ---: | ---: |",
+        ]
+    )
+    for family, stats in sorted(families.items()):
+        lines.append(f"| `{family}` | {stats['total']} | {stats['pass_rate']} |")
+
+    lines.extend(
+        [
+            "",
+            "Full reports: [reports/eval_suite.json](reports/eval_suite.json) and [reports/family_summary.md](reports/family_summary.md)",
+            END_MARKER,
+        ]
+    )
+    return "\n".join(lines)
+
+
+def render_family_summary(report: dict) -> str:
+    summary = report["summary"]
+    lines = [
+        "# Eval Family Summary",
+        "",
+        f"- total cases: `{summary['total_cases']}`",
+        f"- families: `{summary['family_count']}`",
+        f"- false positives: `{summary['false_positives']}`",
+        f"- false negatives: `{summary['false_negatives']}`",
+        f"- average precision: `{summary['average_precision']}`",
+        f"- average recall: `{summary['average_recall']}`",
+        "",
+        "## Family Results",
+        "",
+        "| Family | Cases | Passed | FP | FN | Pass Rate |",
+        "| --- | ---: | ---: | ---: | ---: | ---: |",
+    ]
+    for family, stats in sorted(report["family_summary"].items()):
+        lines.append(
+            f"| `{family}` | {stats['total']} | {stats['passed']} | {stats['false_positives']} | {stats['false_negatives']} | {stats['pass_rate']} |"
+        )
+    return "\n".join(lines) + "\n"
+
+
+def update_readme(readme_path: Path, panel: str) -> None:
+    text = readme_path.read_text(encoding="utf-8")
+    if BEGIN_MARKER not in text or END_MARKER not in text:
+        raise SystemExit(f"README markers not found in {readme_path}")
+    start = text.index(BEGIN_MARKER)
+    end = text.index(END_MARKER) + len(END_MARKER)
+    updated = text[:start] + panel + text[end:]
+    readme_path.write_text(updated, encoding="utf-8")
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Render README-visible eval results and full family report.")
+    parser.add_argument("--eval-dir", default="evals")
+    parser.add_argument("--description-file", default="evals/improved_description.txt")
+    parser.add_argument("--baseline-description-file", default="evals/baseline_description.txt")
+    parser.add_argument("--output-dir", default="reports")
+    parser.add_argument("--readme", default="README.md")
+    args = parser.parse_args()
+
+    report = run_eval_suite(args)
+    output_dir = Path(args.output_dir)
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    (output_dir / "eval_suite.json").write_text(json.dumps(report, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")
+    (output_dir / "family_summary.md").write_text(render_family_summary(report), encoding="utf-8")
+    update_readme(Path(args.readme), render_readme_panel(report))
+
+
+if __name__ == "__main__":
+    main()

scripts/run_eval_suite.py

@@ -33,6 +33,7 @@ def main() -> None:
     parser.add_argument("--eval-dir", default="evals", help="Root eval directory")
     parser.add_argument("--description-file", default="evals/improved_description.txt")
     parser.add_argument("--baseline-description-file", default="evals/baseline_description.txt")
+    parser.add_argument("--output-file", help="Optional path to write the combined JSON report")
     args = parser.parse_args()
 
     root = Path(args.eval_dir).resolve()
@@ -65,15 +66,21 @@ def main() -> None:
     for family, stats in family_summary.items():
         stats["pass_rate"] = round(stats["passed"] / stats["total"], 3) if stats["total"] else None
 
+    total_cases = sum(stats["total"] for stats in family_summary.values())
     summary = {
         "suite_count": len(suites),
+        "total_cases": total_cases,
+        "family_count": len(family_summary),
         "false_positives": aggregate["false_positives"],
         "false_negatives": aggregate["false_negatives"],
         "average_precision": round(sum(aggregate["precision"]) / len(aggregate["precision"]), 3) if aggregate["precision"] else None,
         "average_recall": round(sum(aggregate["recall"]) / len(aggregate["recall"]), 3) if aggregate["recall"] else None,
     }
     output = {"summary": summary, "family_summary": family_summary, "suites": suites}
-    print(json.dumps(output, ensure_ascii=False, indent=2))
+    rendered = json.dumps(output, ensure_ascii=False, indent=2)
+    if args.output_file:
+        Path(args.output_file).write_text(rendered + "\n", encoding="utf-8")
+    print(rendered)
     if summary["false_positives"] > 0 or summary["false_negatives"] > 0:
         raise SystemExit(2)