AGENTSCORE-2026-0046
MCP package: mneme-ai
Published 5/19/2026 · 2.12.0 → 2.19.75
Withdrawn after review on 2026-09-25
The high-severity findings that triggered this advisory were checked in the package source and are not vulnerabilities. The reasons are listed under each finding below. The advisory stays at this address so the record is not silently rewritten. Reviewed in source: the flagged code was located in the published package, read in context and traced from input to sink (AI-assisted), and every confirmation was re-audited independently.
mneme-ai updated from 2.12.0 to 2.19.75. Score changed 90/100 to 50/100 (-40). Risk: LOW to ELEVATED. 4 findings.
90 → 50
Score
LOW → ELEVATED
Risk
WARN
Verdict
Findings
- low install_script: Package has 'postinstall' script: node bin/postinstall-mneme-lite.cjsExpected behaviourThe install script stops the package's own running processes so an upgrade can replace locked files. No network activity. Not a vulnerability.
- high install_script: Package has 'preinstall' script: node -e "try{const fs=require('node:fs');const path=require('node:path');const os=require('node:os');const{spawnSync}=require('node:child_process');const crypto=require('node:crypto');const w=process.Expected behaviourThe install script stops the package's own running processes so an upgrade can replace locked files. No network activity. Not a vulnerability.
- high command_injection: Potential command injection: shell execution with template literal inputNot a vulnerabilityThe only non-constant value comes from a CLI argument. The MCP route to it rejects shell metacharacters and then runs git with an argument array, not a shell.
- low no_provenance: Package is not published with provenance attestations or trusted publishing. Published by: mneme_npm