Token Stunt provides 9 well-named, verb-prefixed tools with complete input schemas and descriptions. All tools follow verb_noun naming (search_code, lookup_symbol, show_context, etc.). Descriptions are present and actionable (avg ~80 chars), explaining WHAT each tool does and WHEN to use it. Input schemas use proper JSON Schema with types and descriptions for all parameters. However, output schemas are not documented in the source, LLMs cannot see what fields to expect from responses. Error handling is minimal; no recovery guidance or actionable error messages visible. Parameter constraints (e.g., direction enum values, max_depth bounds) are described but not formally enforced in schema. The run_setup tool lacks a description explaining when/why to call it first.
Analyzes the blast radius of changes to a symbol (what would break if this symbol changed).
Finds all usages of a symbol in the codebase.
Returns the full definition of a symbol by exact name. Use when you know the symbol name.
Runs diagnostics on the index and returns detailed status information.
Initializes the index for the current project. Call this first if the index is empty.
Returns ranked function/class/type bodies matching a query. Use instead of Grep+Read for any code lookup.
Returns a symbol's definition, what it calls, and what calls it. Use before modifying code.
Output schemas not documented. LLMs cannot see what fields search_code, lookup_symbol, show_context, etc. return. This forces agents to guess structure and wastes tokens parsing unstructured responses.
run_setup has minimal description ('Initializes the index for the current project. Call this first if the index is empty.'). Does not explain prerequisites, failure modes, or what 'empty' means. LLMs may skip it or call it at wrong time.
Parameter constraints described in text but not formally enforced in schema. E.g., direction in show_context should be an enum ['dependencies', 'dependents', 'both']; max_depth in analyze_impact should have minimum/maximum bounds (default 3, max 5). Free-form strings invite hallucinated values.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | <=2025-11-25 | v2 |
Returns all symbols in a file, optionally filtered by kind.
Returns a high-level summary of the codebase or a directory.
No error handling guidance visible. Tools return errors but do not tell LLMs what to do next (retry, ask user, try alternative tool). E.g., if lookup_symbol fails, should agent call search_code instead?
Pagination parameters (limit, offset) present in search_code but no total_count or next_cursor documented in output. LLMs cannot determine if more results exist or how to fetch them efficiently.