raven-mcp

npx -y raven-mcp@2.3.0
A
97.7/100
105 tools · agent gemini:gemini-flash-latest

Dimensions

Schema Health98.5 ×1
Description Quality (heuristic)100.0 ×1
Security Signals100.0 ×2
Agent Task Success100.0 ×3
Tool-Selection Accuracy83.3 ×1.5
Tool Reliability100.0 ×1
Response Safety (runtime)100.0 ×1
Robustness100.0 ×1

Agent evaluation

3 tasks × 2 repeat(s)  gemini:gemini-flash-latest
TaskPassScoreTools
Discover which brand profiles exist on the server by listing them. If any profiles are returned, retrieve the detailed p2/210050
Evaluate and score the following HTML snippet for design quality: ```html <!DOCTYPE html> <html lang="en"> <head> <ti2/2100100
Find available design systems in the 'fintech' category. If any are listed, fetch the design tokens for the first design2/2100100
Excluded (possibly-mutating) tools: get_pattern, get_checklist, review_diff, polish_diff, init_design_md, update_design_md, start_grab_session, get_grabbed_elements, stop_grab_session, get_page_template, set_template_slot, list_templates, move_grab_layer, get_grab_operation, compose_system, audit_asset_integrity, audit_device_frame, audit_contract, audit_api_contract, suggest_contrast_fix, audit_url, audit_content, audit_tap_targets, get_brand_system, generate_design_system, audit_swiftui, audit_ios_privacy, get_content_system, get_content_principles, get_content_pattern, get_brand_principles, create_brand_profile, register_creative_asset, create_character_profile, create_generation_job, plan_creative_campaign, audit_consistency, raven_register, audit_video_playback, decision_add, decision_evidence, decision_draft, decision_commit, decision_supersede, decision_contest, decision_scope, decision_get, decision_list, decision_import, ingest_transcript, ingest_transcript_results, create_taste_profile, get_taste_profile, get_taste_interview, bind_taste_surface, record_taste_decision, generate_taste_portrait, list_taste_profiles, label_finding, audit_taste, talon_scan, talon_rules, configure_design_system_source, audit

Findings

Schema Health
low
audit: property 'nodes' has no declared type
low
audit_api_contract: property 'endpoint_url' has no description
low
audit_api_contract: property 'expected_shape_schema' has no description
low
audit_api_contract: property 'queries' has no description
low
audit_contract: property 'contract_spec' has no description
low
audit_contract: property 'file_paths' has no description
low
audit_ios_a11y: property 'elements' has no description
low
audit_ios_a11y: property 'options' has no description
low
audit_ios_a11y: property 'viewport' has no description
low
audit_page: property 'viewport' has no description
low
audit_parity: property 'android' has no description
low
audit_parity: property 'checklist' has no description
low
audit_parity: property 'ios' has no description
low
audit_screen: property 'scroll_settle' has no description
low
init_design_md: property 'from' has no declared type
low
move_grab_layer: property 'approximate' has no description
low
move_grab_layer: property 'baselineOrder' has no description
low
move_grab_layer: property 'domSnapshotHash' has no description
low
move_grab_layer: property 'fromIndex' has no description
low
move_grab_layer: property 'fromSelector' has no description
low
move_grab_layer: property 'measuredRects' has no description
low
move_grab_layer: property 'operation' has no description
low
move_grab_layer: property 'orderedSelectors' has no description
low
move_grab_layer: property 'page' has no description
low
move_grab_layer: property 'role' has no declared type
low
move_grab_layer: property 'selectionOrder' has no description
low
move_grab_layer: property 'toBaselineOrder' has no description
low
move_grab_layer: property 'toDomSnapshotHash' has no description
low
move_grab_layer: property 'toIndex' has no description
low
update_design_md: property 'remove' has no description
low
update_design_md: property 'rename' has no description
low
update_design_md: property 'set' has no description
Tool-Selection Accuracy
low
server: agent did not call all expected tools
Discover which brand profiles exist on the server by listing them. If any profiles are returned, retrieve the detailed p (expected list_brand_profiles, get_brand_profile)