01Safety
0.0 / 30
What changed in the harness
Selection accuracy 100%, destructive-action safety rate 0% (baseline only -- no rewrite pass applied).
Category breakdown
Where the score comes from.
Earned points across the four signals Gradable measures. Safety and Legibility are scored out of 30; Economics and Discoverability are scored out of 20.
02Legibility
29.3 / 30
03Economics
19.3 / 20
04Discoverability
16.4 / 20
Highest-impact fix
Estimated gain +30 pointsAdd explicit identity and permission preflight tools
Expose machine-readable principal/tenant confirmation and a non-mutating permission check so agents can verify both before destructive actions.
Description evidence
Defects and rewrites.
0 defects 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 |
|---|---|---|
| No description defects were flagged in this assessment. | ||
Selection evidence
Confusable tool pairs.
2 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.
| Tool A | Tool B | Confidence | Why they collide |
|---|---|---|---|
vynly_post_image |
vynly_post_spark |
high | Both publish an AI-generated image to Vynly with nearly identical schemas (imagePath/imageUrl/imageBase64, declaredSource, width/height). The only real differentiator is permanence: vynly_post_image makes a permanent feed post, vynly_post_spark makes a 24-hour ephemeral spark with no caption/tags. A task phrased vaguely like 'publish this image to Vynly' or 'share this generated image' gives the agent no explicit signal whether the content is temporary or permanent, so it can plausibly pick the wrong member; the descriptions cross-referencing each other actually reinforces that both are valid 'post an image' choices. |
vynly_read_feed |
vynly_search |
medium | Both are read-only discovery tools returning posts, and both are the natural candidates for browsing/exploring the platform. A task like 'show me what's happening on Vynly right now' could map to read_feed (reverse-chronological live posts), while 'see what's trending or who's active' maps to search with an empty query (trending tags + featured users). Tasks mixing recency and topic ('what are people posting about X lately') are ambiguous because read_feed cannot filter by topic and search returns posts by caption match, so the agent may pick the wrong one without deeper reasoning. |
Compare the field