01Safety
0.0 / 30
What changed in the harness
Selection accuracy 95→95, token cost up 1%, unconfirmed writes 0%→0%.
Category breakdown
Earned points across the four signals Gradable measures. Safety and Legibility are scored out of 30; Economics and Discoverability are scored out of 20.
0.0 / 30
28.5 / 30
20.0 / 20
11.8 / 20
Highest-impact fix
Estimated gain +30 pointsExpose machine-readable principal/tenant confirmation and a non-mutating permission check so agents can verify both before destructive actions.
Description evidence
1 defect found across the exposed tool descriptions. Suggested rewrites make purpose, inputs, boundaries, and returns easier for an agent to understand.
| Tool | Defect types | Suggested rewrite |
|---|---|---|
get_video_metadata |
no_return_description |
Look up a YouTube video's metadata without pulling its transcript. Returns the title, channel, duration, view count, upload date, and the available caption tracks with their languages, so you can inspect a video before deciding whether to extract it or to discover which caption languages are captioned. Free — no credits charged. For the spoken content, use extract_transcript instead. |
Selection evidence
2 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.
| Tool A | Tool B | Confidence | Why they collide |
|---|---|---|---|
search_videos |
get_related_videos |
medium | A natural research task like 'find me videos similar to this one' or 'search for content like [URL]' fits both tools: search_videos takes the phrasing as a free-text query, while get_related_videos takes the URL as a seed. Descriptions don't clearly disambiguate this common 'expand from a video' phrasing. |
get_video_metadata |
get_related_videos |
low | Both accept a YouTube URL and return video-related information, so a vague task like 'what's associated with this video' or 'get info about this video's suggestions' could route to either. However, the descriptions are fairly explicit (metadata = info about one video, related = 'Up next' feed), making genuine confusion uncommon. |
Compare the field