Loading ToolBench
Your organization's memory layer — an MCP server backed by a git repo.
21 tools with mixed quality. Naming is generally verb-first and clear (read_*, search_*, update_*, add_*, check_*, etc.), following conventions well. However, descriptions vary widely: some are adequate (read_org, search_context), but many are terse or generic (add_info, remove_info, check_roles). Critical gap: NO input schemas are visible in the provided source code. The tools are registered but their actual JSON Schema definitions are not shown, only parameter names and types are inferred from the tool list. This forces a hard cap on schema scores. Output schemas are completely undocumented. Error handling is absent from visible code. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) detected. Composition is reasonable (tools are single-purpose), but parameter descriptions lack detail on constraints, formats, and dependencies.
create a decision record via PR with status:proposed
individual event files - create for significant occurrences
manage context via natural language
agent session records - create at conversation end
team access control - check permissions
team access control - check roles
team access control - configure CODEOWNERS
get full content of a memory entry
No input schemas visible in source code. Tool definitions show parameter names and types but no JSON Schema with type constraints, enums, or format declarations. Hard cap on schema scoring.
Output schemas completely undocumented. No visible documentation of what fields tools return, their types, or structure. LLMs cannot plan downstream calls or extract data reliably.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 41 | 2026-07-28+ | v2 |
get full map of decisions, sessions, and events
get inbound and outbound links for a memory entry
bring .txt, .md, or .docx files into the knowledge base
list all section headings in organisation.md
propose updates via pull requests for review
view the full organisation.md file
read a specific section by heading
manage context via natural language
create GitHub issues automatically
find information across org file, context files, and all memory stores. Supports filtering by status, date, type
create GitHub issues automatically
change decision status (proposed → accepted | superseded). Only status is editable — decision bodies are immutable
propose updates via pull requests for review
Parameter descriptions lack constraint details. 'heading' parameter has no guidance on format or valid values. 'status' in search_context has no enum. 'rules' in configure_codeowners has no structure documentation.
No error handling guidance visible. Tools lack recovery hints, retryability classification, or actionable error messages. Agents cannot self-correct on failures.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) present. Agents cannot distinguish safe reads from destructive writes without parsing descriptions.