Loading ToolBench
US H-1B visa sponsorship, LCA wages & top sponsors by role — recruiting/talent data
Three well-defined tools with clear, domain-specific descriptions (194 - 280 chars each, within baseline 34 - 392 range). All tools have proper JSON Schema with typed parameters and descriptions. Naming follows verb_noun pattern (h1b_employer_sponsorship, h1b_salary, h1b_top_sponsors). However, output schemas are not documented, LLMs cannot predict return structure. No error handling guidance. Parameters lack constraints (enums, ranges, patterns) despite accepting free-form strings (employer, job_title, city). No pagination guidance despite tools returning samples/lists. Missing tool annotations (readOnlyHint, idempotentHint). All three tools are READ_ONLY and idempotent, but this is not declared in schema.
Check whether a US employer sponsors H-1B visas and profile their sponsorship, from DOL Labor Condition Application (LCA) disclosures. Answers 'does company X sponsor H-1B / green cards' for recruiting and candidate advising. Returns the number of certified LCA filings, base-salary range (min / median / max), the top sponsored job titles, and top work locations for the employer. Filter by year (defaults to the latest full year). Employer name is matched as the disclosed legal name (e.g. 'Google', 'Amazon.com Services').
Look up real H-1B base salaries for a job title from DOL LCA disclosures — a market wage benchmark backed by actual filed salaries (not estimates). Answers 'what do H-1B software engineers earn at company X / in city Y'. Filter by job title, and optionally by employer, city, and year. Returns salary statistics (count, min / median / average / max) plus a sample of individual records (employer, title, salary, location, dates).
Find which US employers sponsor the most H-1B visas for a given job title (optionally in a specific city) — a candidate-sourcing / target-account signal for recruiting. Answers 'which companies sponsor the most data engineers in Austin' or 'top H-1B sponsors for nurses'. Returns employers ranked by certified LCA filings for the role, each with their filing count and median base salary. Backed by real DOL LCA disclosures.
Output schemas not documented. LLMs cannot predict return structure (salary stats, employer profiles, ranked lists). Forces agents to guess field names and types.
Parameters accept free-form strings without constraints. 'employer', 'job_title', 'city' lack enums, patterns, or format hints. LLMs may pass invalid values (e.g. 'SEATTLE' vs 'Seattle') causing silent mismatches.
No error handling guidance. If upstream h1bdata.info is down or returns no results, LLMs receive raw errors with no recovery path. Missing 'retryable' vs 'user-fixable' classification.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 61 | 2026-07-28+ | v2 |
Tool annotations missing. All three tools are READ_ONLY and idempotent, but inputSchema lacks readOnlyHint and idempotentHint. Agents cannot infer safety properties.
Pagination not addressed. h1b_salary and h1b_top_sponsors accept 'limit' but no offset/cursor. Code caps results at 20,000 rows internally but does not expose pagination to LLM. Large result sets risk context window exhaustion.