Multi-tier web search & browser automation for AI agents. SearXNG + Camofox + CloakBrowser.
Critical gaps across naming, descriptions, and schemas. Tool names lack consistent verb prefixes (health, getTabs, snapshot lack action clarity). Descriptions are present but minimal (10-50 chars for most tools), providing insufficient context for LLM selection. Input schemas exist but lack depth: most parameters have only type and description, no enums for constrained inputs (e.g., 'direction' in scroll, 'format' in cloak-fetch). Output schemas are entirely undocumented. Error handling guidance is absent. The cloak-fetch tool has rich parameter documentation but stands alone; 18 other tools are under-specified. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications. Parameter descriptions are generic ('Tab identifier', 'URL to navigate to') without format constraints or examples. No pagination guidance for tools returning lists.
Click an element in a tab
Fetch any URL through CloakBrowser stealth Chromium with anti-detection capabilities
Close a browser tab
Create a new browser tab with specified URL
Destroy the current session
Detect anti-bot challenges (Cloudflare, Akamai, DataDome, Imperva, PerimeterX, DDoS-Guard)
Execute JavaScript expression in a tab
Extract structured data from a tab using a schema
Inconsistent verb-based naming. Tools like 'health', 'getTabs', 'snapshot', 'screenshotRaw' lack clear action verbs. LLMs struggle to infer intent from non-verb names. Should be 'check_health', 'list_tabs', 'get_snapshot', 'take_screenshot'.
Descriptions are too brief (10-50 chars for 17/19 tools). Baseline for A+ tools is 50-200 chars. Current descriptions lack context for when/why to use each tool. E.g., 'Check Camofox browser health status' does not explain what 'health' means or what to do if unhealthy.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields. E.g., getTabs returns a list but structure is unknown. cloak-fetch returns content but format is undocumented.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 40 | 2026-07-28+ | v2 |
Get list of open tabs
Check Camofox browser health status
Navigate a tab to a new URL
Extract article content from a tab using Readability.js
Take a screenshot of a tab
Scroll a tab in specified direction
Get accessibility tree snapshot of a tab
Start the browser
Stop the browser
Type text into an element in a tab
Wait for an element to appear in a tab
Missing tool annotations despite clear risk classifications. Tools marked WRITE (start, stop, createTab, closeTab, evaluate, click, type, scroll, navigate, destroySession) lack destructiveHint. Tools marked READ_ONLY lack readOnlyHint. Idempotent operations (createTab with same URL) lack idempotentHint.
Parameter descriptions lack format constraints and enums. E.g., 'direction' in scroll has no enum (up/down/left/right?). 'format' in cloak-fetch correctly uses enum but is exception. 'ref' in click/type lacks explanation of reference format. No validation guidance for LLMs.