Loading ToolBench
Adaptive agent team orchestration for Claude Code — agents, skills, commands, rules, workflows, and hooks
This server has severe definition quality issues. 6 of 8 tools lack visible input schemas in the source code provided. Descriptions exist but are often vague or lack actionable guidance. Tool names are inconsistent (verb_noun vs noun patterns). No output schemas documented. Error handling is absent. The codebase appears to be shell scripts and hooks rather than a properly structured MCP server with formal tool registration.
Runs before git commit. Scans tracking files for incomplete items. Warns (does not block) if items are incomplete — the orchestrator should see this warning and address remaining items.
Stop Hook: Check for console.log statements in modified files. Cross-platform (Windows, macOS, Linux). Runs after each response and checks if any modified JavaScript/TypeScript files contain console.log statements. Provides warnings to help developers remember to remove debug statements before committing.
Checks if a hook is enabled based on hook ID and profiles. Returns 'yes' or 'no'.
Cost Tracker Hook. Appends lightweight session usage metrics to ~/.claude/metrics/costs.jsonl. Estimates token costs based on model type and token counts.
Dispatch Skill Reminder Hook (PreToolUse: Agent). Fires before the Agent tool is invoked. Scans plan documents for a Skills Found table and prints a reminder to include skill text in the agent dispatch prompt.
Doc file warning hook (PreToolUse - Write). Warns about non-standard documentation files. Exit code 0 always (warns only, never blocks).
No input schemas visible for 6 of 8 tools (batch-completion-check, check-console-log, dispatch-skill-reminder, and others). Cannot verify parameter types, constraints, or descriptions.
Tool names lack consistent verb_noun pattern. 'batch-completion-check', 'check-console-log', 'check-hook-enabled' use check_ prefix but 'dispatch-skill-reminder' and 'doc-file-warning' do not. LLMs struggle to infer intent from inconsistent naming.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields. E.g., what does 'check-hook-enabled' return, a boolean, a string, an object?
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 34 | 2026-07-28+ | v2 |
Continuous Learning v2 - Observer background loop. Analyzes observations and creates instinct files. Includes re-entrancy guard, cooldown throttle, and tail-based sampling to prevent memory explosion from runaway parallel Claude analysis processes.
Starts the background observer agent that analyzes observations and creates instincts. Uses Haiku model for cost efficiency. Project-scoped — detects current project and analyzes project-specific observations into project-scoped instincts.
Descriptions are generic and lack actionable guidance. E.g., 'Checks if a hook is enabled' does not explain WHEN to call this tool or what to do with the result. Descriptions should be 50-200 chars and answer: what, when, why.
No error handling or recovery guidance. Tools do not document what happens on failure, whether errors are retryable, or what the LLM should do next. E.g., observer-loop can timeout or fail, no guidance provided.