Two tools with explicit schemas, descriptions, and tool annotations. Both tools have well-structured input/output schemas using Zod with proper type definitions and parameter descriptions. Tool names are action-oriented (create, docs_context). Descriptions are substantive (100-200 chars). However, parameter descriptions contain template variables (${FLOW_STACKS}, ${DEFAULT_STACK}) that are not expanded in the visible code, reducing clarity for LLMs. Output schemas are documented. Error handling is basic, no recovery guidance or actionable error messages visible in the tool implementations.
MemoryBank one-shot initialize/create flow. Returns executable instructions for building a project-specific memory bank from real codebase patterns.
Search official documentation context for a target stack and return compact snippets (title, url, excerpt, score). Use detailLevel=structured when you need richer extracted context for top results; details may include summary, declarations, code examples, and deprecation signals.
Parameter descriptions use unexpanded template variables (${FLOW_STACKS}, ${DEFAULT_STACK}, ${DOCS_CONTEXT_DETAIL_LEVELS}). LLMs cannot resolve these at runtime and will see literal template syntax instead of actual enum values.
Error handling in tool implementations returns generic error messages without recovery guidance. Example: 'Invalid arguments for tool create: ...' does not suggest next steps or valid alternatives.
docs_context tool accepts 'queries' as required array parameter but no guidance on query format, length limits, or examples. Parameter description says 'Prefer specific API names' but does not explain what happens with vague queries.
create tool's 'stack' parameter is optional with no default documented in the description. The code references DEFAULT_AGENT_PROVIDER but the parameter description does not explain what happens if stack is omitted.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 68 | 2026-07-28+ | v2 |
No pagination support visible in docs_context despite returning potentially large result sets. Tool accepts maxItemsPerQuery (max 10) but no cursor/offset for iterating beyond top results.