Bird Recognition Tool

Problem

Bird recognition in field workflows often fails on narrow but practical details: the target bird is not clearly framed, similar species or subspecies are hard to separate, and a one-off label is not enough for later review. This skill is aimed at ecological observation, backyard birding, and bird diversity surveys, where the output needs to be a structured, re-queryable report rather than a temporary classification.

How It Works

The input supports images, video, local files, and network URLs. It runs scripts.bird_recognition_analysis against the cloud analysis service to perform bird detection, species classification, and confidence scoring. A typical flow is:

• Provide a clear clip where the bird is visible, with formats such as jpg/jpeg/png/mp4/avi/mov and a 10MB size limit.
• Run the script from the skill root, passing either a local path or a URL; URLs are downloaded by the API rather than locally.
• Return a structured analysis report with findings, suggestions, and a report link; historical reports can be queried through --list and rendered as a Markdown table.

Limits

The skill is best treated as an observation aid. Its output is useful for reference but should not replace expert verification, local regulations, or legal and scientific workflows. For video, use a stable, unobstructed segment; blur, occlusion, or long distance can reduce classification reliability.