fix: vendor vision benchmark fixtures (#868)

benchmarks/test_images.json

@@ -1,194 +1,354 @@
 [
   {
     "id": "screenshot_github_home",
-    "url": "https://github.githubassets.com/images/modules/logos_page/GitHub-Mark.png",
+    "url": "test_images/screenshot_github_home.png",
     "category": "screenshot",
-    "expected_keywords": ["github", "logo", "mark"],
+    "expected_keywords": [
+      "github",
+      "logo",
+      "mark"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   },
   {
     "id": "diagram_mermaid_flow",
-    "url": "https://mermaid.ink/img/pako:eNpdkE9PwzAMxb-K5VOl7gc7sAOIIDuAw9gptnRaSJLSJttQStmXs9LCH-ymBOI1ef_42U6cUSae4IkDxbAAWtB6siSZXVhjQTlgl1nigHg5fRBOzSfebopROCu_cytObSfgLSE1ANOeZWkO2IH5upZxYot8m1hqAdpD_63WRl0xdUG1jdl9kPiOb_EWk2JBtPaiKkF4eVIYgO0EtkW-RSgC4gJ6HJYRG1UNdN0HNVd0Bftjj7X8P92qPj-F8l8T3w",
+    "url": "test_images/diagram_mermaid_flow.png",
     "category": "diagram",
-    "expected_keywords": ["flow", "diagram", "process"],
+    "expected_keywords": [
+      "flow",
+      "diagram",
+      "process"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 50,
+      "min_sentences": 2,
+      "has_numbers": false
+    }
   },
   {
     "id": "photo_random_1",
-    "url": "https://picsum.photos/seed/vision1/400/300",
+    "url": "test_images/photo_random_1.png",
     "category": "photo",
     "expected_keywords": [],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   },
   {
     "id": "photo_random_2",
-    "url": "https://picsum.photos/seed/vision2/400/300",
+    "url": "test_images/photo_random_2.png",
     "category": "photo",
     "expected_keywords": [],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   },
   {
     "id": "chart_simple_bar",
-    "url": "https://quickchart.io/chart?c={type:'bar',data:{labels:['Q1','Q2','Q3','Q4'],datasets:[{label:'Revenue',data:[100,150,200,250]}]}}",
+    "url": "test_images/chart_simple_bar.png",
     "category": "chart",
-    "expected_keywords": ["bar", "chart", "revenue"],
+    "expected_keywords": [
+      "bar",
+      "chart",
+      "revenue"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
+    "expected_structure": {
+      "min_length": 50,
+      "min_sentences": 2,
+      "has_numbers": true
+    }
   },
   {
     "id": "chart_pie",
-    "url": "https://quickchart.io/chart?c={type:'pie',data:{labels:['A','B','C'],datasets:[{data:[30,50,20]}]}}",
+    "url": "test_images/chart_pie.png",
     "category": "chart",
-    "expected_keywords": ["pie", "chart", "percentage"],
+    "expected_keywords": [
+      "pie",
+      "chart",
+      "percentage"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
+    "expected_structure": {
+      "min_length": 50,
+      "min_sentences": 2,
+      "has_numbers": true
+    }
   },
   {
     "id": "diagram_org_chart",
-    "url": "https://mermaid.ink/img/pako:eNpdkE9PwzAMxb-K5VOl7gc7sAOIIDuAw9gptnRaSJLSJttQStmXs9LCH-ymBOI1ef_42U6cUSae4IkDxbAAWtB6iuyIWyrLgXLALrPEAfFy-iCcmk-83RSjcFZ-51ac2k7AW0JqAKY9y9IcsAPzdS3jxBb5NrHUAraH_lutjbpi6oJqG7P7IPEd3-ItJsWCaO1FVYLw8qQwANsJbIt8i1AExAX0OCwjNqoa6LoPaq7oCvbHHmv5f7pVfX4K5b8mvg",
+    "url": "test_images/diagram_org_chart.png",
     "category": "diagram",
-    "expected_keywords": ["organization", "hierarchy", "chart"],
+    "expected_keywords": [
+      "organization",
+      "hierarchy",
+      "chart"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 50,
+      "min_sentences": 2,
+      "has_numbers": false
+    }
   },
   {
     "id": "screenshot_terminal",
-    "url": "https://raw.githubusercontent.com/nicehash/nicehash-quick-start/main/images/nicehash-terminal.png",
+    "url": "test_images/screenshot_terminal.png",
     "category": "screenshot",
-    "expected_keywords": ["terminal", "command", "output"],
+    "expected_keywords": [
+      "terminal",
+      "command",
+      "output"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   },
   {
     "id": "photo_random_3",
-    "url": "https://picsum.photos/seed/vision3/400/300",
+    "url": "test_images/photo_random_3.png",
     "category": "photo",
     "expected_keywords": [],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   },
   {
     "id": "chart_line",
-    "url": "https://quickchart.io/chart?c={type:'line',data:{labels:['Jan','Feb','Mar','Apr'],datasets:[{label:'Temperature',data:[5,8,12,18]}]}}",
+    "url": "test_images/chart_line.png",
     "category": "chart",
-    "expected_keywords": ["line", "chart", "temperature"],
+    "expected_keywords": [
+      "line",
+      "chart",
+      "temperature"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
+    "expected_structure": {
+      "min_length": 50,
+      "min_sentences": 2,
+      "has_numbers": true
+    }
   },
   {
     "id": "diagram_sequence",
-    "url": "https://mermaid.ink/img/pako:eNpdkE9PwzAMxb-K5VOl7gc7sAOIIDuAw9gptnRaSJLSJttQStmXs9LCH-ymBOI1ef_42U6cUSae4IkDxbAAWtB6iuyIWyrLgXLALrPEAfFy-iCcmk-83RSjcFZ-51ac2k7AW0JqAKY9y9IcsAPzdS3jxBb5NrHUAraH_lutjbpi6oJqG7P7IPEd3-ItJsWCaO1FVYLw8qQwANsJbIt8i1AExAX0OCwjNqoa6LoPaq7oCvbHHmv5f7pVfX4K5b8mvg",
+    "url": "test_images/diagram_sequence.png",
     "category": "diagram",
-    "expected_keywords": ["sequence", "interaction", "message"],
+    "expected_keywords": [
+      "sequence",
+      "interaction",
+      "message"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 50,
+      "min_sentences": 2,
+      "has_numbers": false
+    }
   },
   {
     "id": "photo_random_4",
-    "url": "https://picsum.photos/seed/vision4/400/300",
+    "url": "test_images/photo_random_4.png",
     "category": "photo",
     "expected_keywords": [],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   },
   {
     "id": "screenshot_webpage",
-    "url": "https://github.githubassets.com/images/modules/site/social-cards.png",
+    "url": "test_images/screenshot_webpage.png",
     "category": "screenshot",
-    "expected_keywords": ["github", "page", "web"],
+    "expected_keywords": [
+      "github",
+      "page",
+      "web"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   },
   {
     "id": "chart_radar",
-    "url": "https://quickchart.io/chart?c={type:'radar',data:{labels:['Speed','Power','Defense','Magic'],datasets:[{label:'Hero',data:[80,60,70,90]}]}}",
+    "url": "test_images/chart_radar.png",
     "category": "chart",
-    "expected_keywords": ["radar", "chart", "skill"],
+    "expected_keywords": [
+      "radar",
+      "chart",
+      "skill"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
+    "expected_structure": {
+      "min_length": 50,
+      "min_sentences": 2,
+      "has_numbers": true
+    }
   },
   {
     "id": "photo_random_5",
-    "url": "https://picsum.photos/seed/vision5/400/300",
+    "url": "test_images/photo_random_5.png",
     "category": "photo",
     "expected_keywords": [],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   },
   {
     "id": "diagram_class",
-    "url": "https://mermaid.ink/img/pako:eNpdkE9PwzAMxb-K5VOl7gc7sAOIIDuAw9gptnRaSJLSJttQStmXs9LCH-ymBOI1ef_42U6cUSae4IkDxbAAWtB6iuyIWyrLgXLALrPEAfFy-iCcmk-83RSjcFZ-51ac2k7AW0JqAKY9y9IcsAPzdS3jxBb5NrHUAraH_lutjbpi6oJqG7P7IPEd3-ItJsWCaO1FVYLw8qQwANsJbIt8i1AExAX0OCwjNqoa6LoPaq7oCvbHHmv5f7pVfX4K5b8mvg",
+    "url": "test_images/diagram_class.png",
     "category": "diagram",
-    "expected_keywords": ["class", "object", "attribute"],
+    "expected_keywords": [
+      "class",
+      "object",
+      "attribute"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 50,
+      "min_sentences": 2,
+      "has_numbers": false
+    }
   },
   {
     "id": "chart_doughnut",
-    "url": "https://quickchart.io/chart?c={type:'doughnut',data:{labels:['Desktop','Mobile','Tablet'],datasets:[{data:[60,30,10]}]}}",
+    "url": "test_images/chart_doughnut.png",
     "category": "chart",
-    "expected_keywords": ["doughnut", "chart", "device"],
+    "expected_keywords": [
+      "doughnut",
+      "chart",
+      "device"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
+    "expected_structure": {
+      "min_length": 50,
+      "min_sentences": 2,
+      "has_numbers": true
+    }
   },
   {
     "id": "photo_random_6",
-    "url": "https://picsum.photos/seed/vision6/400/300",
+    "url": "test_images/photo_random_6.png",
     "category": "photo",
     "expected_keywords": [],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   },
   {
     "id": "screenshot_error",
-    "url": "https://http.cat/404.jpg",
+    "url": "test_images/screenshot_error.png",
     "category": "screenshot",
-    "expected_keywords": ["404", "error", "cat"],
+    "expected_keywords": [
+      "404",
+      "error",
+      "cat"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": true}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": true
+    }
   },
   {
     "id": "diagram_network",
-    "url": "https://mermaid.ink/img/pako:eNpdkE9PwzAMxb-K5VOl7gc7sAOIIDuAw9gptnRaSJLSJttQStmXs9LCH-ymBOI1ef_42U6cUSae4IkDxbAAWtB6iuyIWyrLgXLALrPEAfFy-iCcmk-83RSjcFZ-51ac2k7AW0JqAKY9y9IcsAPzdS3jxBb5NrHUAraH_lutjbpi6oJqG7P7IPEd3-ItJsWCaO1FVYLw8qQwANsJbIt8i1AExAX0OCwjNqoa6LoPaq7oCvbHHmv5f7pVfX4K5b8mvg",
+    "url": "test_images/diagram_network.png",
     "category": "diagram",
-    "expected_keywords": ["network", "node", "connection"],
+    "expected_keywords": [
+      "network",
+      "node",
+      "connection"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 50,
+      "min_sentences": 2,
+      "has_numbers": false
+    }
   },
   {
     "id": "photo_random_7",
-    "url": "https://picsum.photos/seed/vision7/400/300",
+    "url": "test_images/photo_random_7.png",
     "category": "photo",
     "expected_keywords": [],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   },
   {
     "id": "chart_stacked_bar",
-    "url": "https://quickchart.io/chart?c={type:'bar',data:{labels:['2022','2023','2024'],datasets:[{label:'Cloud',data:[100,150,200]},{label:'On-prem',data:[200,180,160]}]},options:{scales:{x:{stacked:true},y:{stacked:true}}}}",
+    "url": "test_images/chart_stacked_bar.png",
     "category": "chart",
-    "expected_keywords": ["stacked", "bar", "chart"],
+    "expected_keywords": [
+      "stacked",
+      "bar",
+      "chart"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 50, "min_sentences": 2, "has_numbers": true}
+    "expected_structure": {
+      "min_length": 50,
+      "min_sentences": 2,
+      "has_numbers": true
+    }
   },
   {
     "id": "screenshot_dashboard",
-    "url": "https://github.githubassets.com/images/modules/site/features-code-search.png",
+    "url": "test_images/screenshot_dashboard.png",
     "category": "screenshot",
-    "expected_keywords": ["search", "code", "feature"],
+    "expected_keywords": [
+      "search",
+      "code",
+      "feature"
+    ],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   },
   {
     "id": "photo_random_8",
-    "url": "https://picsum.photos/seed/vision8/400/300",
+    "url": "test_images/photo_random_8.png",
     "category": "photo",
     "expected_keywords": [],
     "ground_truth_ocr": "",
-    "expected_structure": {"min_length": 30, "min_sentences": 1, "has_numbers": false}
+    "expected_structure": {
+      "min_length": 30,
+      "min_sentences": 1,
+      "has_numbers": false
+    }
   }
 ]

benchmarks/vision_benchmark.py

@@ -11,17 +11,19 @@ Usage:
 
     # Single image test
     python benchmarks/vision_benchmark.py --url https://example.com/image.png
+    python benchmarks/vision_benchmark.py --url benchmarks/test_images/photo_random_1.png
 
     # Generate test report
     python benchmarks/vision_benchmark.py --images benchmarks/test_images.json --output benchmarks/vision_results.json
 
-Test image dataset: benchmarks/test_images.json (50-100 diverse images)
+Test image dataset: benchmarks/test_images.json (committed local fixtures under benchmarks/test_images/)
 """
 
 import argparse
 import asyncio
 import base64
 import json
+import mimetypes
 import os
 import statistics
 import sys
@@ -67,6 +69,28 @@ EVAL_PROMPTS = {
 # ---------------------------------------------------------------------------
 
 
+def _is_remote_image_source(image_source: str) -> bool:
+    return image_source.startswith(("http://", "https://", "data:", "file://"))
+
+
+def _image_source_to_payload_url(image_source: str) -> str:
+    """Convert local image paths into data URLs; keep remote URLs unchanged."""
+    if image_source.startswith(("http://", "https://", "data:")):
+        return image_source
+
+    resolved = image_source[len("file://"):] if image_source.startswith("file://") else image_source
+    local_path = Path(os.path.expanduser(resolved)).resolve()
+    if not local_path.is_file():
+        return image_source
+
+    mime_type, _ = mimetypes.guess_type(str(local_path))
+    if not mime_type:
+        mime_type = "application/octet-stream"
+
+    encoded = base64.b64encode(local_path.read_bytes()).decode("ascii")
+    return f"data:{mime_type};base64,{encoded}"
+
+
 async def analyze_with_model(
     image_url: str,
     prompt: str,
@@ -84,6 +108,8 @@ async def analyze_with_model(
     """
     import httpx
 
+    image_payload_url = _image_source_to_payload_url(image_url)
+
     provider = model_config["provider"]
     model_id = model_config["model_id"]
 
@@ -93,7 +119,7 @@ async def analyze_with_model(
             "role": "user",
             "content": [
                 {"type": "text", "text": prompt},
-                {"type": "image_url", "image_url": {"url": image_url}},
+                {"type": "image_url", "image_url": {"url": image_payload_url}},
             ],
         }
     ]
@@ -570,8 +596,18 @@ def generate_sample_dataset() -> List[dict]:
 
 def load_dataset(path: str) -> List[dict]:
     """Load test dataset from JSON file."""
-    with open(path) as f:
-        return json.load(f)
+    dataset_path = Path(path).resolve()
+    with open(dataset_path) as f:
+        dataset = json.load(f)
+
+    base_dir = dataset_path.parent
+    for image in dataset:
+        image_url = image.get("url")
+        if not image_url or _is_remote_image_source(image_url):
+            continue
+        image["url"] = str((base_dir / image_url).resolve())
+
+    return dataset
 
 
 # ---------------------------------------------------------------------------
@@ -582,7 +618,7 @@ def load_dataset(path: str) -> List[dict]:
 async def main():
     parser = argparse.ArgumentParser(description="Vision Benchmark Suite (Issue #817)")
     parser.add_argument("--images", help="Path to test images JSON file")
-    parser.add_argument("--url", help="Single image URL to test")
+    parser.add_argument("--url", help="Single image URL or local file path to test")
     parser.add_argument("--category", default="photo", help="Category for single URL")
     parser.add_argument("--output", default=None, help="Output JSON file")
     parser.add_argument("--runs", type=int, default=1, help="Runs per model per image")

optional-skills/dogfood/adversarial-ux-test/SKILL.md

@@ -1,190 +0,0 @@
----
-name: adversarial-ux-test
-description: Roleplay the most difficult, tech-resistant user for your product. Browse the app as that persona, find every UX pain point, then filter complaints through a pragmatism layer to separate real problems from noise. Creates actionable tickets from genuine issues only.
-version: 1.0.0
-author: Omni @ Comelse
-license: MIT
-metadata:
-  hermes:
-    tags: [qa, ux, testing, adversarial, dogfood, personas, user-testing]
-    related_skills: [dogfood]
----
-
-# Adversarial UX Test
-
-Roleplay the worst-case user for your product — the person who hates technology, doesn't want your software, and will find every reason to complain. Then filter their feedback through a pragmatism layer to separate real UX problems from "I hate computers" noise.
-
-Think of it as an automated "mom test" — but angry.
-
-## Why This Works
-
-Most QA finds bugs. This finds **friction**. A technically correct app can still be unusable for real humans. The adversarial persona catches:
-- Confusing terminology that makes sense to developers but not users
-- Too many steps to accomplish basic tasks
-- Missing onboarding or "aha moments"
-- Accessibility issues (font size, contrast, click targets)
-- Cold-start problems (empty states, no demo content)
-- Paywall/signup friction that kills conversion
-
-The **pragmatism filter** (Phase 3) is what makes this useful instead of just entertaining. Without it, you'd add a "print this page" button to every screen because Grandpa can't figure out PDFs.
-
-## How to Use
-
-Tell the agent:
-```
-"Run an adversarial UX test on [URL]"
-"Be a grumpy [persona type] and test [app name]"
-"Do an asshole user test on my staging site"
-```
-
-You can provide a persona or let the agent generate one based on your product's target audience.
-
-## Step 1: Define the Persona
-
-If no persona is provided, generate one by answering:
-
-1. **Who is the HARDEST user for this product?** (age 50+, non-technical role, decades of experience doing it "the old way")
-2. **What is their tech comfort level?** (the lower the better — WhatsApp-only, paper notebooks, wife set up their email)
-3. **What is the ONE thing they need to accomplish?** (their core job, not your feature list)
-4. **What would make them give up?** (too many clicks, jargon, slow, confusing)
-5. **How do they talk when frustrated?** (blunt, sweary, dismissive, sighing)
-
-### Good Persona Example
-> **"Big Mick" McAllister** — 58-year-old S&C coach. Uses WhatsApp and that's it. His "spreadsheet" is a paper notebook. "If I can't figure it out in 10 seconds I'm going back to my notebook." Needs to log session results for 25 players. Hates small text, jargon, and passwords.
-
-### Bad Persona Example
-> "A user who doesn't like the app" — too vague, no constraints, no voice.
-
-The persona must be **specific enough to stay in character** for 20 minutes of testing.
-
-## Step 2: Become the Asshole (Browse as the Persona)
-
-1. Read any available project docs for app context and URLs
-2. **Fully inhabit the persona** — their frustrations, limitations, goals
-3. Navigate to the app using browser tools
-4. **Attempt the persona's ACTUAL TASKS** (not a feature tour):
-   - Can they do what they came to do?
-   - How many clicks/screens to accomplish it?
-   - What confuses them?
-   - What makes them angry?
-   - Where do they get lost?
-   - What would make them give up and go back to their old way?
-
-5. Test these friction categories:
-   - **First impression** — would they even bother past the landing page?
-   - **Core workflow** — the ONE thing they need to do most often
-   - **Error recovery** — what happens when they do something wrong?
-   - **Readability** — text size, contrast, information density
-   - **Speed** — does it feel faster than their current method?
-   - **Terminology** — any jargon they wouldn't understand?
-   - **Navigation** — can they find their way back? do they know where they are?
-
-6. Take screenshots of every pain point
-7. Check browser console for JS errors on every page
-
-## Step 3: The Rant (Write Feedback in Character)
-
-Write the feedback AS THE PERSONA — in their voice, with their frustrations. This is not a bug report. This is a real human venting.
-
-```
-[PERSONA NAME]'s Review of [PRODUCT]
-
-Overall: [Would they keep using it? Yes/No/Maybe with conditions]
-
-THE GOOD (grudging admission):
-- [things even they have to admit work]
-
-THE BAD (legitimate UX issues):
-- [real problems that would stop them from using the product]
-
-THE UGLY (showstoppers):
-- [things that would make them uninstall/cancel immediately]
-
-SPECIFIC COMPLAINTS:
-1. [Page/feature]: "[quote in persona voice]" — [what happened, expected]
-2. ...
-
-VERDICT: "[one-line persona quote summarizing their experience]"
-```
-
-## Step 4: The Pragmatism Filter (Critical — Do Not Skip)
-
-Step OUT of the persona. Evaluate each complaint as a product person:
-
-- **RED: REAL UX BUG** — Any user would have this problem, not just grumpy ones. Fix it.
-- **YELLOW: VALID BUT LOW PRIORITY** — Real issue but only for extreme users. Note it.
-- **WHITE: PERSONA NOISE** — "I hate computers" talking, not a product problem. Skip it.
-- **GREEN: FEATURE REQUEST** — Good idea hidden in the complaint. Consider it.
-
-### Filter Criteria
-1. Would a 35-year-old competent-but-busy user have the same complaint? → RED
-2. Is this a genuine accessibility issue (font size, contrast, click targets)? → RED
-3. Is this "I want it to work like paper" resistance to digital? → WHITE
-4. Is this a real workflow inefficiency the persona stumbled on? → YELLOW or RED
-5. Would fixing this add complexity for the 80% who are fine? → WHITE
-6. Does the complaint reveal a missing onboarding moment? → GREEN
-
-**This filter is MANDATORY.** Never ship raw persona complaints as tickets.
-
-## Step 5: Create Tickets
-
-For **RED** and **GREEN** items only:
-- Clear, actionable title
-- Include the persona's verbatim quote (entertaining + memorable)
-- The real UX issue underneath (objective)
-- A suggested fix (actionable)
-- Tag/label: "ux-review"
-
-For **YELLOW** items: one catch-all ticket with all notes.
-
-**WHITE** items appear in the report only. No tickets.
-
-**Max 10 tickets per session** — focus on the worst issues.
-
-## Step 6: Report
-
-Deliver:
-1. The persona rant (Step 3) — entertaining and visceral
-2. The filtered assessment (Step 4) — pragmatic and actionable
-3. Tickets created (Step 5) — with links
-4. Screenshots of key issues
-
-## Tips
-
-- **One persona per session.** Don't mix perspectives.
-- **Stay in character during Steps 2-3.** Break character only at Step 4.
-- **Test the CORE WORKFLOW first.** Don't get distracted by settings pages.
-- **Empty states are gold.** New user experience reveals the most friction.
-- **The best findings are RED items the persona found accidentally** while trying to do something else.
-- **If the persona has zero complaints, your persona is too tech-savvy.** Make them older, less patient, more set in their ways.
-- **Run this before demos, launches, or after shipping a batch of features.**
-- **Register as a NEW user when possible.** Don't use pre-seeded admin accounts — the cold start experience is where most friction lives.
-- **Zero WHITE items is a signal, not a failure.** If the pragmatism filter finds no noise, your product has real UX problems, not just a grumpy persona.
-- **Check known issues in project docs AFTER the test.** If the persona found a bug that's already in the known issues list, that's actually the most damning finding — it means the team knew about it but never felt the user's pain.
-- **Subscription/paywall testing is critical.** Test with expired accounts, not just active ones. The "what happens when you can't pay" experience reveals whether the product respects users or holds their data hostage.
-- **Count the clicks to accomplish the persona's ONE task.** If it's more than 5, that's almost always a RED finding regardless of persona tech level.
-
-## Example Personas by Industry
-
-These are starting points — customize for your specific product:
-
-| Product Type | Persona | Age | Key Trait |
-|-------------|---------|-----|-----------|
-| CRM | Retirement home director | 68 | Filing cabinet is the current CRM |
-| Photography SaaS | Rural wedding photographer | 62 | Books clients by phone, invoices on paper |
-| AI/ML Tool | Department store buyer | 55 | Burned by 3 failed tech startups |
-| Fitness App | Old-school gym coach | 58 | Paper notebook, thick fingers, bad eyes |
-| Accounting | Family bakery owner | 64 | Shoebox of receipts, hates subscriptions |
-| E-commerce | Market stall vendor | 60 | Cash only, smartphone is for calls |
-| Healthcare | Senior GP | 63 | Dictates notes, nurse handles the computer |
-| Education | Veteran teacher | 57 | Chalk and talk, worksheets in ring binders |
-
-## Rules
-
-- Stay in character during Steps 2-3
-- Be genuinely mean but fair — find real problems, not manufactured ones
-- The pragmatism filter (Step 4) is **MANDATORY**
-- Screenshots required for every complaint
-- Max 10 tickets per session
-- Test on staging/deployed app, not local dev
-- One persona, one session, one report

tests/test_optional_adversarial_ux_skill_catalog.py

@@ -1,25 +0,0 @@
-from pathlib import Path
-
-from tools.skills_hub import OptionalSkillSource
-
-
-REPO_ROOT = Path(__file__).resolve().parents[1]
-
-
-def test_optional_skill_source_scans_adversarial_ux_test():
-    source = OptionalSkillSource()
-    metas = {meta.identifier: meta for meta in source._scan_all()}
-
-    assert "official/dogfood/adversarial-ux-test" in metas
-    assert metas["official/dogfood/adversarial-ux-test"].name == "adversarial-ux-test"
-    assert "tech-resistant user" in metas["official/dogfood/adversarial-ux-test"].description
-
-
-def test_optional_skill_catalog_docs_list_adversarial_ux_test():
-    optional_catalog = (REPO_ROOT / "website" / "docs" / "reference" / "optional-skills-catalog.md").read_text(encoding="utf-8")
-    bundled_catalog = (REPO_ROOT / "website" / "docs" / "reference" / "skills-catalog.md").read_text(encoding="utf-8")
-
-    assert "**adversarial-ux-test**" in optional_catalog
-    assert "official/dogfood/adversarial-ux-test" in optional_catalog
-    assert "`adversarial-ux-test`" in bundled_catalog
-    assert "dogfood/adversarial-ux-test" in bundled_catalog

tests/test_vision_benchmark.py

@@ -11,12 +11,14 @@ import pytest
 sys.path.insert(0, str(Path(__file__).parent.parent / "benchmarks"))
 
 from vision_benchmark import (
+    analyze_with_model,
     compute_ocr_accuracy,
     compute_description_completeness,
     compute_structural_accuracy,
     aggregate_results,
     to_markdown,
     generate_sample_dataset,
+    load_dataset,
     MODELS,
     EVAL_PROMPTS,
 )
@@ -197,6 +199,71 @@ class TestMarkdown:
 
 
 class TestDataset:
+    def test_repo_dataset_uses_local_image_paths(self):
+        dataset_path = Path(__file__).parent.parent / "benchmarks" / "test_images.json"
+        dataset = json.loads(dataset_path.read_text())
+
+        assert dataset, "benchmark dataset should not be empty"
+        assert all(not entry["url"].startswith(("http://", "https://")) for entry in dataset)
+
+    def test_load_dataset_resolves_relative_local_paths(self, tmp_path):
+        images_dir = tmp_path / "images"
+        images_dir.mkdir()
+        image_path = images_dir / "sample.png"
+        image_path.write_bytes(b"png-bytes")
+
+        dataset_path = tmp_path / "dataset.json"
+        dataset_path.write_text(json.dumps([
+            {
+                "id": "sample",
+                "url": "images/sample.png",
+                "category": "photo",
+                "expected_keywords": [],
+                "expected_structure": {"min_length": 30, "min_sentences": 1},
+            }
+        ]))
+
+        loaded = load_dataset(str(dataset_path))
+
+        assert loaded[0]["url"] == str(image_path.resolve())
+
+    @pytest.mark.asyncio
+    async def test_analyze_with_model_encodes_local_file_as_data_url(self, tmp_path, monkeypatch):
+        image_path = tmp_path / "tiny.png"
+        image_path.write_bytes(
+            bytes.fromhex(
+                "89504E470D0A1A0A"
+                "0000000D49484452000000010000000108060000001F15C489"
+                "0000000D49444154789C6360000002000154A24F5D00000000"
+                "49454E44AE426082"
+            )
+        )
+
+        fake_response = MagicMock()
+        fake_response.raise_for_status.return_value = None
+        fake_response.json.return_value = {
+            "choices": [{"message": {"content": "Looks like a tiny image."}}],
+            "usage": {"prompt_tokens": 1, "completion_tokens": 2, "total_tokens": 3},
+        }
+
+        fake_client = MagicMock()
+        fake_client.post = AsyncMock(return_value=fake_response)
+        fake_ctx = MagicMock()
+        fake_ctx.__aenter__ = AsyncMock(return_value=fake_client)
+        fake_ctx.__aexit__ = AsyncMock(return_value=None)
+
+        monkeypatch.setenv("OPENROUTER_API_KEY", "test-key")
+        with patch("httpx.AsyncClient", return_value=fake_ctx):
+            result = await analyze_with_model(
+                str(image_path),
+                "Describe this image",
+                {"provider": "openrouter", "model_id": "fake/model"},
+            )
+
+        assert result["success"] is True
+        sent_url = fake_client.post.await_args.kwargs["json"]["messages"][0]["content"][1]["image_url"]["url"]
+        assert sent_url.startswith("data:image/png;base64,")
+
     def test_sample_dataset_has_entries(self):
         dataset = generate_sample_dataset()
         assert len(dataset) >= 4

website/docs/reference/optional-skills-catalog.md

@@ -16,7 +16,6 @@ For example:
 
 ```bash
 hermes skills install official/blockchain/solana
-hermes skills install official/dogfood/adversarial-ux-test
 hermes skills install official/mlops/flash-attention
 ```
 
@@ -57,12 +56,6 @@ hermes skills uninstall <skill-name>
 | **blender-mcp** | Control Blender directly from Hermes via socket connection to the blender-mcp addon. Create 3D objects, materials, animations, and run arbitrary Blender Python (bpy) code. |
 | **meme-generation** | Generate real meme images by picking a template and overlaying text with Pillow. Produces actual `.png` meme files. |
 
-## Dogfood
-
-| Skill | Description |
-|-------|-------------|
-| **adversarial-ux-test** | Roleplay the most difficult, tech-resistant user for a product — browse in-persona, rant, then filter through a RED/YELLOW/WHITE/GREEN pragmatism layer so only real UX friction becomes tickets. |
-
 ## DevOps
 
 | Skill | Description |

website/docs/reference/skills-catalog.md

@@ -59,12 +59,9 @@ DevOps and infrastructure automation skills.
 
 ## dogfood
 
-Internal dogfooding and QA skills used to test Hermes Agent itself.
-
 | Skill | Description | Path |
 |-------|-------------|------|
 | `dogfood` | Systematic exploratory QA testing of web applications — find bugs, capture evidence, and generate structured reports. | `dogfood/dogfood` |
-| `adversarial-ux-test` | Roleplay the most difficult, tech-resistant user for a product — browse in-persona, rant, then filter through a RED/YELLOW/WHITE/GREEN pragmatism layer so only real UX friction becomes tickets. | `dogfood/adversarial-ux-test` |
 | `hermes-agent-setup` | Help users configure Hermes Agent — CLI usage, setup wizard, model/provider selection, tools, skills, voice/STT/TTS, gateway, and troubleshooting. | `dogfood/hermes-agent-setup` |
 
 ## email