Loading ToolBench
Server has 30 tools with consistent naming (verb_noun pattern: blumira_navigate, blumira_findings_list, etc.) and structured schemas. However, descriptions are generic and lack LLM-optimization guidance. Most parameters have types but lack depth. Output schemas are not documented. Error handling is minimal. Tool composition is reasonable (findings, agents, users, msp, resolutions domains) but lacks idempotency guarantees and recovery guidance.
Get an agent device by ID.
List agent devices in the organization.
Get an agent key by ID.
List agent keys in the organization.
Return to the domain navigation menu.
Assign owners to a finding. Owner types: responder, analyst, manager. Empty owners array clears assignments.
Output schemas not documented. LLMs cannot plan downstream calls or extract required fields (e.g., finding_id from list results). Baseline: 100% of A+ tools document return types.
Generic descriptions lack LLM-optimization guidance. Many descriptions under 100 chars (e.g., 'List agent devices in the organization') do not explain WHEN to use the tool or what distinguishes it from similar tools. Baseline: A+ tools average 194 chars with context.
No error recovery guidance. Tools return errors but do not guide LLM on next steps (e.g., 'Invalid finding_id. Call blumira_findings_list to discover valid IDs'). Baseline: A+ tools include recovery hints.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
Add a comment to a finding. Body may contain HTML.
List comments on a finding.
Get detailed finding info including owners, resolution, category, summary, and UI URL.
Get a finding by ID.
List findings with optional filtering by status, priority, category, date ranges, and name patterns.
Resolve a finding. Resolution IDs: 10=Valid, 20=False Positive, 30=No Action Needed, 40=Risk Accepted.
Get MSP account details including license, agent counts, and user count.
List MSP sub-accounts with names and open finding counts.
Get an agent device from an MSP account.
List agent devices for an MSP account.
List findings across all MSP accounts.
Assign owners to a finding in an MSP account.
Add a comment to a finding in an MSP account.
List comments on a finding in an MSP account.
Get a specific finding from an MSP account.
List findings for a specific MSP account.
Resolve a finding in an MSP account.
Get an agent key from an MSP account.
List agent keys for an MSP account.
List users for an MSP account.
Navigate to a domain to see its tools. Domains: findings, agents, users, msp, resolutions. - findings: list/search, get, get details, resolve, assign owners, comments - agents: list devices, get device, list keys, get key - users: list organization users - msp: MSP multi-account management — accounts, per-account findings/agents/users - resolutions: list available resolution options
List available resolution options for findings (e.g., 10=Valid, 20=False Positive, 30=No Action Needed, 40=Risk Accepted).
Check Blumira API connection status and available domains.
List users in the organization. Returns user IDs, emails, names, and roles.
No idempotency guarantees documented. Destructive tools (resolve, assign, add_comment) do not state whether repeated calls with same input are safe. Agents retry on ambiguous failures, non-idempotent tools risk duplicate side effects.
Pagination parameters (page, page_size, limit) lack constraints. No min/max documented. Unbounded numbers let LLMs pass absurd values (e.g., page_size=999999) that break APIs or cause timeouts.