Resolves an agent's request into exact documented capabilities from a recursive tree, with Claude and Codex harnesses to compare it against skills.
Single tool with minimal schema and generic description. Input schema present but lacks parameter descriptions. Tool description is adequate (89 chars) but parameter 'request' has no description in the schema itself. No output schema documented. No error handling guidance. No tool annotations. Composition is monolithic, the tool attempts to resolve capabilities into 'exact documented capability items' but the mechanism is opaque to the LLM.
Resolve a complete request into minimal context for exact documented capability items such as commands, procedures, and references. It never executes them.
Parameter 'request' lacks description in schema. LLM cannot infer what constitutes a 'complete request' or what 'every dependent step' means.
No output schema documented. LLM does not know what fields to expect from the resolution result, making downstream tool chaining impossible.
No error handling guidance. If resolution fails (malformed request, missing capabilities directory), the LLM receives only a generic error message with no recovery path.
Tool name 'resolve_capabilities' is vague. Does it return a list of capabilities, a single matched capability, or a structured plan? Verb is clear but object is ambiguous.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 52 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint, idempotentHint). LLM cannot determine if this tool is safe to retry or has side effects.