Plan-and-execute AI coding agent with verification and safe tool orchestration
RxyCode exposes 30 tools with minimal schema documentation. Most tools lack visible input parameter schemas in the provided source code. Descriptions are present but generic (10-50 chars typical). No evidence of structured output schemas, error handling guidance, or parameter type definitions. Tool names follow verb_noun convention (read, write, edit, bash, grep, glob, ls, view, webfetch, websearch, git, datetime, history, question, skill, change_directory, diagnostics, format, memory, vision, workflow, task, patch, open_file, download_skill, download_mcp, team_install, file_download, task_manage, subagent_task) but descriptions are minimal. Only 3 tools (datetime, download_skill, download_mcp) show explicit parameter schemas in the source; the rest are inferred from file references without visible schema definitions. No evidence of output schema documentation, pagination support, or error recovery guidance.
Execute bash commands
Change the current working directory
Get current system date and time. Use this when user asks about today's date, current time, or datetime.
Run system diagnostics
Add or remove an MCP server configuration.
Install a skill from GitHub or a direct URL, or remove an installed skill.
Edit file contents with line-based modifications
Download files
27 of 30 tools lack visible input parameter schemas. Only datetime, download_skill, and download_mcp show explicit schema definitions. Per HARD SCORING RULE, tools without visible schemas score 0 on schema dimension.
Descriptions are generic and under 50 characters for 27 tools (e.g., 'Read file contents', 'Write content to a file'). Current descriptions lack WHEN to use, prerequisites, or state-modification warnings.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 40 | 2026-07-28+ | v2 |
Format code or text
Execute git commands
Find files matching a pattern
Search for patterns in files
Access command history
List directory contents
Manage memory and context
Open a file in the editor
Apply patches to files
Ask the user a question
Read file contents
Manage skills
Dispatch tasks to subagents
Manage tasks and subagent dispatch
Manage task list
Install team configurations
View file contents with syntax highlighting
Process images and visual content
Fetch content from a URL
Search the web using DuckDuckGo
Manage workflows
Write content to a file
No evidence of output schema documentation across any tool. LLMs cannot plan downstream calls or extract required fields without knowing response structure. Missing pagination support (limit, offset, total_count) for list-returning tools (ls, glob, grep, history).
Destructive tools (bash, git, write, edit, patch, team_install, download_mcp) lack error handling guidance, confirmation patterns, or dry-run support. No evidence of recovery guides or error categorization (retryable vs fatal).
No parameter descriptions visible for any tool except datetime, download_skill, download_mcp. Tools like bash, edit, patch, and git accept complex inputs but provide no guidance on valid values, formats, or constraints. LLMs cannot infer parameter semantics from names alone.