01Safety
0.0 / 30
What changed in the harness
Selection accuracy 100→94, token cost up 1%, 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
28.8 / 30
03Economics
19.3 / 20
04Discoverability
14.2 / 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.
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 |
|---|---|---|
impri_report_result |
no_return_description |
Report whether you successfully executed an approved action. Closes the audit loop — the operator sees 'executed' or 'execute_failed' in the inbox alongside the original action and decision. Always call this after attempting an approved action, even on failure. Statuses: "executed" — action was carried out successfully "execute_failed" — execution attempt failed (include the error in detail) Returns a confirmation (e.g. { accepted: true }) once the result has been recorded for the action. |
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 |
|---|---|---|---|
impri_create_watcher |
impri_create_watcher_from_preset |
high | Both tools create watchers for the same sources (RSS, Reddit, GitHub); a task like 'set up a watcher for the MachineLearning subreddit' matches both — one builds the spec manually, the other uses a preset, and nothing in the request signals which path to take. |
impri_list_watchers |
impri_list_watcher_presets |
medium | Names share 'list watcher'; a vague task like 'list all watchers' or 'what watchers are available to me' is ambiguous between configured watchers and preset templates, since 'available' could mean either. |
impri_await_decision |
impri_inbox_status |
medium | Both surface pending human-approval state; a task like 'check on pending approvals' or 'is anything waiting on a human decision' could map to polling a specific action for its decision or to querying overall queue depth/titles, especially when no action_id is mentioned. |
Compare the field