Loading ToolBench
MCP server for Life Context Vault, a local-first life context management system that provides policy-filtered context packs and memory proposals to AI clients
Life Context Vault MCP defines 4 tools with reasonable naming (verb_noun pattern) and descriptions. However, critical gaps exist: parameter descriptions are sparse or missing entirely, output schemas are not documented, and error handling guidance is absent. The tools operate on sensitive personal data but lack security-focused descriptions (e.g., no mention of confirmation flows, data retention, or privacy guarantees). Naming is clear (request_context_pack, propose_memory, get_policy_summary, get_request_status) but descriptions are brief (50-100 chars) and lack WHEN/WHY context. Parameters like 'sensitivityCeiling' lack detailed format guidance. No documented output schemas means LLMs cannot plan downstream operations or extract required fields.
Return the configured connector and policy summary without exposing raw Vault contents.
Check whether a queued Context Pack request has been confirmed, denied, expired, or is still pending.
Create an unapproved Memory Inbox candidate from the current conversation. This never creates an ApprovedFact.
Request a short-lived, policy-filtered Context Pack. Sensitive packs are queued for user confirmation instead of returned directly.
Output schemas not documented. LLMs cannot infer what fields are returned, breaking downstream tool chaining and forcing discovery calls.
Parameter descriptions missing or incomplete. 'sensitivityCeiling' enum is defined but lacks explanation of what each tier means or when to use each. 'taskText' and 'clientName' lack guidance on format/length constraints.
No error handling guidance. Tools operating on sensitive personal data should document failure modes (e.g., 'request denied by user', 'context pack expired') and recovery steps, but none are provided.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
Descriptions lack WHEN/WHY context. 'Request a short-lived, policy-filtered Context Pack' does not explain when an LLM should call this vs. propose_memory, or what 'policy-filtered' means in practice.
Security descriptions absent. Tools handle sensitive personal data but descriptions do not mention confirmation flows, data retention, or privacy guarantees. Users need to know what happens to proposed memories and how long context packs persist.