AI Skill Report Card
Searching Yandex Images
YAML--- name: searching-yandex-images description: Searches Yandex Images to find, filter, and refine visual results by size, orientation, type, color, and format, or to locate the source/context of an image. Use when asked to find images via Yandex, do a reverse image search, or narrow image results using visual filters. --- # Searching Yandex Images
Quick Start13 / 15
- Go to
yandex.com/images(or the regional domain, e.g.yandex.ru/images). - Enter a text query in the search bar, or click the camera icon ("Search by image") to upload a photo / paste an image URL / drag-and-drop a file.
- If using image search, use the crop frame handles to select the specific region of the photo you want Yandex to focus on (e.g. isolate a face, object, or logo).
- Apply filters (Size, Orientation, Type, Color, Format, Recent, On the site) to narrow results.
- Scan results; click "On the site" filter or open an image to jump to its source page for context/attribution.
Recommendation▾
Examples are somewhat abstract (describe steps rather than concrete simulated outputs like actual result counts, filenames, or URLs found) — add a realistic example with mock search results.
Workflow13 / 15
Progress:
- Step 1: Determine search mode — text query vs. reverse image search
- Step 2: Enter query or upload image
- Step 3: (Reverse search only) Adjust selection frame to target the relevant subject
- Step 4: Apply filters to reduce noise
- Step 5: Review results and open source pages as needed
- Step 6: Refine query/filters if results are off-target
Step 1 — Choose mode:
- Text query: best for general/conceptual searches ("red vintage bicycle").
- Reverse image search (Smart Camera): best for finding the origin of an image, similar images, identifying objects/people/places, or finding higher-resolution versions.
Step 2 — Input:
- Text: type keywords, be as specific as possible (subject + attributes + context).
- Image: upload file, paste URL, or drag image onto the search box.
Step 3 — Refine selection (reverse search only):
- Yandex auto-detects likely subjects and draws a frame; drag frame edges to isolate the exact area of interest (crops out background clutter).
- Re-run search after adjusting frame if the initial detection was too broad or too narrow.
Step 4 — Apply filters:
- Size: small / medium / large / wallpaper-size, or exact pixel dimensions.
- Orientation: horizontal, vertical, square.
- Type: photo, clipart, line drawing, face, etc.
- Color: full-color, black-and-white, or a specific dominant color.
- Format: JPEG, PNG, GIF, etc.
- Recent: restrict to recently indexed/published images.
- On the site: restrict results to a specific domain.
Step 5 — Review & verify:
- Click through to source pages to confirm context, licensing, and authenticity — especially important for reverse image searches used to fact-check or trace origin.
Step 6 — Refine:
- If results are too broad: add filters or more specific keywords.
- If results are too narrow/empty: remove a filter or broaden the crop frame.
Recommendation▾
Address edge cases like Yandex requiring a browser vs API-less access for automated agents, region blocks, or CAPTCHA challenges which are likely real pitfalls for a Claude skill.
Examples13 / 20
Example 1: Input: Find the original source of a photo you found on social media. Output: Click Smart Camera → upload the photo → adjust frame to isolate the main subject (ignore watermarks/borders) → run search → open "On the site" results to identify earliest/original publisher.
Example 2: Input: Find high-resolution square wallpapers of mountain landscapes. Output: Text query "mountain landscape" → apply Size filter "wallpaper" or set custom large dimensions → Orientation: square → Type: photo → review results.
Example 3: Input: Identify what object or logo appears in a cropped photo. Output: Upload image via Smart Camera → drag frame tightly around just the object/logo → search → check "On the site" links and similar-image matches to identify it.
Recommendation▾
Since this is likely used programmatically, clarify whether this assumes browser automation (e.g., via a tool) and give selector/URL specifics instead of generic UI descriptions.
Best Practices
- Tighten the crop frame around only the subject of interest before running a reverse search — background elements dilute match accuracy.
- Combine filters rather than relying on one; e.g., Type + Color + Orientation together yield much cleaner results than keywords alone.
- Use "On the site" to verify provenance before trusting or reusing an image.
- For fact-checking, cross-check multiple source pages returned by reverse search, not just the first result.
- Use specific, descriptive keywords (subject + style + color + context) rather than single generic words.
Common Pitfalls
- Uploading a full uncropped image when only one element (a face, product, logo) matters — this returns noisy, irrelevant matches.
- Ignoring available filters and manually scrolling through hundreds of unfiltered results.
- Treating the first result as ground truth for image origin without checking multiple source links.
- Forgetting that filters like Size/Format persist across new searches — clear or check them if results seem unexpectedly limited.