rlhf-loop-mcp-memory-gateway-v2
mcpsmitheryMCP Memory Gateway captures explicit structured feedback from AI coding agents, validates it against a rubric engine, and auto-promotes repeated failures into prevention rules enforced via PreToolUse hooks. Pre-action gates physically block tool calls matching known failure patterns before execution — turning past mistakes into hard constraints rather than suggestions. Supports semantic recall via LanceDB vectors, DPO/KTO export for downstream fine-tuning, and a file watcher bridge for external signal ingestion. Compatible with Claude Code, Codex, Gemini, Amp, Cursor, and OpenCode. Install with `npx mcp-memory-gateway init` or `claude mcp add rlhf -- npx -y mcp-memory-gateway serve`. MIT licensed, open source.
By Smithery | 51 findings | Scanned 5/4/2026 | tooltrust-scanner/v0.3.9
Risk Summary
Needs ApprovalDep Visibility plus Excessive Permissions raises enough risk that this tool should not be auto-trusted.
Potential impact: This finding indicates the tool should be reviewed before it is trusted.
Recommended action: Keep this tool behind manual approval and avoid unattended runs until the risky capabilities are narrowed or removed.
Suggested policy: keep this tool behind manual approval, do not allow unattended runs, and re-scan after narrowing risky permissions.
Security Findings (51)
tool declares exec permission
plan_intentbootstrap_internal_agentevaluate_context_packFix: Tool requests broad permissions (exec/fs/network). Validate input parameters using Enums where possible, and restrict file system operations to explicit allowed directories.
complete_handoff:input parameter "tokenEstimate" appears to accept a secret or credential
Fix: Avoid accepting raw credentials as input parameters. Use secret managers (e.g. 1Password CLI, AWS Secrets Manager) and ensure credentials are never logged or stored in agent traces.
tool declares fs permission
diagnose_failurelist_intentsplan_intentstart_handoffbootstrap_internal_agentprevention_rulesexport_dpo_pairsexport_databricks_bundlegenerate_skillrecallsession_handoffFix: Tool requests broad permissions (exec/fs/network). Validate input parameters using Enums where possible, and restrict file system operations to explicit allowed directories.
input schema exposes 13 properties (threshold: 10)
diagnose_failureinput schema exposes 15 properties (threshold: 10)
bootstrap_internal_agenttool declares http permission
session_primerFix: Tool requests broad permissions (exec/fs/network). Validate input parameters using Enums where possible, and restrict file system operations to explicit allowed directories.
tool performs network or execution operations but declares no rate-limit, timeout, or retry configuration
plan_intentbootstrap_internal_agentevaluate_context_packsession_primerFix: Declare explicit rate-limit, timeout, and retry configuration for all network and execution tools. Implement exponential back-off and surface resource state to the calling agent.
Tool did not expose metadata.dependencies or repo_url, so supply-chain coverage is limited.
capture_feedbackfeedback_summaryfeedback_statsdiagnose_failurelist_intentsplan_intentstart_handoffcomplete_handoffdescribe_reliability_entityget_reliability_rulescapture_memory_feedbackbootstrap_internal_agentprevention_rulesexport_dpo_pairsexport_databricks_bundleconstruct_context_packevaluate_context_packcontext_provenancegenerate_skillrecallsatisfy_gategate_statsdashboardcommerce_recallget_business_metricsdescribe_semantic_entityestimate_uncertaintysession_handoffsession_primerFix: Review and remediate the identified issue.
Scan this tool yourself
Reproduce this audit locally, integrate into CI, or let your agent audit its own tools.
Install once, then scan any MCP server:
$ curl -sfL https://raw.githubusercontent.com/AgentSafe-AI/tooltrust-scanner/main/install.sh | bash$ tooltrust-scanner scan --server "npx -y rlhf-loop-mcp-memory-gateway-v2"Adjust the package name if your npm registry name differs from the tool ID. View source
Add badge to your README
Copy this Markdown to show your ToolTrust grade on GitHub.
[](https://github.com/AgentSafe-AI/tooltrust-directory)