@minimax-ai/code@0.5.5
⚠ Under reviewMinimax Code — terminal coding agent
AI Security Review
scanned 1mo ago · by lpm-firewall-aiThe postinstall hook and the two agent-configuration hints behave like a local SQLite install check and normal coding-agent file reads, with no self-install chain. A clean decision still cannot be issued because twelve required citation paths exceed the eight-citation cap, so the review stays unresolved.
Decision evidence
public snapshot- package.json defines one lifecycle script, postinstall, which runs node ./verify-native-install.mjs.
- That verifier, when launched as the process entry, checks the Node and glibc range and smoke-tests optional better-sqlite3 with an in-memory query.
- Declared dependencies are @mariozechner/clipboard and @vscode/ripgrep, not this package name.
- In chunks/chunk-S6GSNWAM.js the names AGENTS.md and CLAUDE.md are members of a read-tool filename set that uses readFile and access.
- The Codex helper in chunks/launcher-GHPADSKI.js joins CODEX_HOME or ~/.codex/sessions and returns if those directory reads fail.
- Twelve paths are required for a clean citation record and this review allows at most eight citations, so a mark_clean result cannot be finalized.
- The postinstall path prints a compatibility result or an install hint and does not write agent configuration or transmit credentials.
- The signed updater downloads a release and runs npm install only inside an explicit apply path after signature and checksum checks.
Source & flagged code
23 flaggedPackage defines install-time lifecycle scripts.
package.jsonView on unpkgPackage source references child process execution.
chunks/launcher-GHPADSKI.jsView on unpkg · L9Findings
Affected versions and remediation
This report applies to @minimax-ai/code@0.5.5.
See all versions of @minimax-ai/code for other recorded verdicts.
- Review the evidence and your use of @minimax-ai/code@0.5.5 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.