Add vision-analysis skill; move workaround from TOOLS.md; update log-work-expense to skip image analysis

vision-analysis/SKILL.md

@@ -0,0 +1,22 @@
+---
+name: vision-analysis
+description: Analyze images using a vision model via Python. Use when the built-in image tool fails or when you need to extract text, describe content, or analyze one or more images. Triggers on any request to analyze, read, describe, or extract information from images.
+---
+
+# Vision Analysis
+
+Analyze images by calling the OpenRouter API directly via Python.
+
+## Script
+
+```bash
+uv run --with requests ~/notes/skills/vision-analysis/scripts/analyze.py "<prompt>" <image1> [image2] ...
+```
+
+- **Prompt** (required): What to extract or analyze from the images.
+- **Images** (required): One or more local file paths.
+
+## Notes
+
+- The built-in `image` tool does not work with the configured vision model. Always use this script instead.
+- Max ~20 images per call (API limit).

vision-analysis/scripts/analyze.py

@@ -0,0 +1,20 @@
+import requests, json, sys, base64
+
+API_KEY = "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"
+MODEL = "google/gemini-2.5-flash-lite"
+
+prompt = sys.argv[1]
+paths = sys.argv[2:]
+
+content = [{"type": "text", "text": prompt}]
+for p in paths:
+    with open(p, "rb") as f:
+        b64 = base64.b64encode(f.read()).decode()
+    content.append({"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{b64}"}})
+
+resp = requests.post(
+    "https://openrouter.ai/api/v1/chat/completions",
+    headers={"Authorization": f"Bearer {API_KEY}", "Content-Type": "application/json"},
+    json={"model": MODEL, "messages": [{"role": "user", "content": content}], "max_tokens": 2000}
+)
+print(resp.json()["choices"][0]["message"]["content"])