Loading ToolBench
The open, model-neutral context and session framework for coding agents.
ScoutCTX defines 2 tools with complete input schemas and reasonable descriptions. Both tools follow verb_noun naming (scout_context, scout_session_context) and accept structured parameters with type constraints. However, descriptions lack depth on WHEN to use each tool vs. the other, and output schemas are not documented. Parameter descriptions are present but minimal (10-30 chars), below the 72-char baseline. No error handling guidance or recovery paths documented. Tools are read-only (low risk) but lack idempotent/destructive annotations.
Build a focused, token-budgeted, secret-redacted context package for an AI coding task.
Rebuild context for a persistent ScoutCTX session, including its durable task, notes, harness history, and current worktree.
Output schemas not documented. LLMs cannot plan downstream calls or extract fields from Brief/ContextResult objects. Rubric requires 100% of A+ tools document return types.
Parameter descriptions are minimal (10-30 chars vs. 72-char baseline). E.g., 'budget' says 'Approximate maximum output token count' but does not explain how it interacts with max_files or what happens if budget is exhausted.
No distinction between scout_context and scout_session_context in descriptions. LLMs cannot determine when to use one vs. the other. Tool descriptions must answer 'When should the LLM call it instead of a similar tool?'
No error handling guidance. If session_id is invalid or root directory does not exist, what should the LLM do? Descriptions lack recovery paths (e.g., 'If session not found, call scout_context instead').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2025-06-18+ | v2 |
Tool annotations missing. Both tools are read-only and idempotent, but readOnlyHint and idempotentHint are not declared in schemas. This prevents clients from optimizing caching or retry logic.