01Safety
0.0 / 30
Public leaderboard
Public assessment
AutomateLab-tech/citation-intelligence (@automatelab/citation-intelligence)
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
27.7 / 30
03Economics
8.5 / 20
04Discoverability
12.2 / 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.
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.
8 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.
| Tool A | Tool B | Confidence | Why they collide |
|---|---|---|---|
audit_sitemap |
audit_sitemap_map |
high | Both accept a sitemap_url and are labelled sitemap audits; the descriptions even call sitemap_map the 'inverse' of audit_sitemap. A task like 'audit my sitemap to see which pages get cited' routes to either: audit_sitemap scores citability, audit_sitemap_map reports actual cached citations. |
domain_am_i_cited |
domain_cited_for |
medium | Both operate on a domain and return per-query citation presence, so tasks like 'is my domain cited' or 'what queries is my domain cited for' are ambiguous: am_i_cited needs the queries supplied, while cited_for lists queries from cache. |
domain_cited_for |
domain_cited_for_diff |
medium | Both take a domain and return per-query citation lists; a task like 'show me the queries I'm cited for, gained or lost' could land on plain cited_for instead of the time-window diff (or the diff instead of the full list) given only the descriptions. |
signals_gsc_gap |
signals_bing_gap |
medium | Descriptions are near-identical (join search-console rank with am_i_cited to surface rank-but-not-cited queries). A task 'find queries I rank well for but am not cited in' doesn't name Google vs Bing, so either tool satisfies it. |
citations_check |
citations_evidence |
medium | Both run check_citations on a query and return the cited URLs with raw-answer handling; evidence just adds snippet extraction. 'Show me the citations/evidence an AI gives for X' fits both since evidence also returns the full citation list. |
citations_check |
citations_provenance |
medium | Both fan a query and return cited URLs; provenance is check_citations fanned across engines for cross-engine consensus. A task like 'which sources does AI cite for X, especially in common across engines' is ambiguous between the single-engine and fan-out tools. |
competitors_compare |
competitors_compete |
medium | Both benchmark a URL against competitors; compare takes explicit urls, compete takes a query + your_url and runs the pipeline end-to-end. 'Compare my URL against top cited competitors' is ambiguous when the user doesn't state whether they have URLs or a query. |
panel_track |
panel_run |
medium | Both address named panels with overlapping input (name, domain, engine). A user saying 'track this panel' might mean save/load it (panel_track) or execute a monitoring run that snapshots citations (panel_run), since the verb 'track' in the request maps to the wrong tool. |
Compare the field