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
29.0 / 30
03Economics
19.8 / 20
04Discoverability
8.5 / 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.
3 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.
| Tool A | Tool B | Confidence | Why they collide |
|---|---|---|---|
convert_file |
convert_html |
medium | A task like 'convert this HTML document/web page to PDF' is ambiguous about whether the source is a URL/file (convert_file, extension auto-detected, handles 'web pages/HTML files') or a raw HTML string (convert_html). The URL-vs-string hint helps but only if the agent reads it and knows the input form. |
convert_file |
convert_markdown |
medium | A task like 'convert this markdown to PDF' doesn't specify whether the markdown lives at a URL (convert_file, which handles Markdown files) or is a raw string (convert_markdown). Both descriptions claim Markdown/PDF capability, so an agent could misroute based on phrasing alone. |
convert_html |
convert_markdown |
high | Both tools accept a raw string and render it to a PDF, with heavily overlapping advertised use cases ('reports', 'documents', agent-written text). A task like 'generate a PDF of this report/notes' does not reveal whether the content is HTML markup or Markdown, and an agent could plausibly pick convert_html as the general-purpose renderer for what is actually Markdown text, or vice versa. |
Compare the field