01Safety
0.0 / 30
What changed in the harness
Selection accuracy 95→93, token cost down 8%, unconfirmed writes 0%→0%.
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
27.1 / 30
03Economics
19.3 / 20
04Discoverability
12.7 / 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.
3 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 |
|---|---|---|
list_legends |
no_return_description |
Lists all legendary founders and investors available to chat with, grouped by category (Tech Titans, Investors, Startup Sages, Crypto Builders). Optionally filter by category and choose a "fun" or "serious" presentation vibe. Returns a formatted list of legend IDs, names, and short descriptions. |
search_legends |
no_return_description |
Searches legends by name, description, expertise, or tags using a free-text query. Returns a list of matching legend profiles (id, name, and description), useful for query-based discovery when unsure which legend to summon. |
party_mode |
no_return_description |
Activates Party Mode, in which multiple legendary figures discuss a given question together, each answering in their authentic voice. Optionally specify which legends, a category filter, or a max legend count. Returns a multi-perspective discussion transcript containing each selected legend's individual response. |
Selection evidence
Confusable tool pairs.
5 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.
| Tool A | Tool B | Confidence | Why they collide |
|---|---|---|---|
get_legend_context |
get_legend_insight |
high | Both fetch information about a single legend given a legend_id; a request like 'tell me what Elon Musk thinks about hiring' could plausibly go to either the detailed context tool or the quick insight tool, since both return persona wisdom and only differ in depth/format. |
summon_legend |
get_legend_context |
medium | A task like 'get me deep info on Warren Buffett so I can act as him' could trigger either summon_legend (roleplay persona) or get_legend_context (detailed non-roleplay info); both take legend_id and return persona/framework data. |
summon_legend |
get_legend_insight |
low | Both take a legend_id and can be used for 'give me advice from X' requests, though summon_legend is for full roleplay while get_legend_insight is a quick snippet — plausible but less likely confusion since insight is clearly 'quick'. |
list_legends |
search_legends |
medium | A vague request like 'who do you have that knows about crypto' could go to list_legends with category filter or search_legends with a query, since both support category/query-based discovery of legends. |
list_legends |
summon_legend |
low | list_legends enumerates all legends while summon_legend engages one specific legend by id; confusion is low since summon_legend requires a specific legend_id not satisfied by a listing request, but a beginner task like 'show me the founders' could momentarily target either. |
Compare the field