hippo-memory@1.37.0
Biologically-inspired memory system for AI agents. Decay by default, strength through use.
AI Security Review
scanned 14d ago · by lpm-firewall-aiThe install hook conditionally restores a same-vendor Codex wrapper based on existing integration metadata, which is a guarded lifecycle integration that merits a warning. The inspected network code sends embeddings to the selected provider; the source does not establish credential theft or another active attack.
Decision evidence
public snapshot- The manifest runs a postinstall script when the package is installed.
- The postinstall path calls a function that repairs the Codex wrapper only when saved integration metadata exists.
- The wrapper repair uses paths under the user's Hippo directory.
- The capture helper reads Claude project session files, and the CLI lists agent configuration files for setup detection.
- The hook declarations expose an explicit OpenCode plugin install function.
- Embedding requests use the configured provider URL and send embedding text with the provider API key.
- The postinstall code does not create a new Codex wrapper on first install; it requires existing metadata.
- The inspected provider code sends requests to the selected embedding provider endpoint.
Source & flagged code
11 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgPackage source references a known benign dynamic code generation pattern.
dist/rerankers/cross-encoder.jsView on unpkg · L9Findings
Affected versions and remediation
This report applies to hippo-memory@1.37.0.
See all versions of hippo-memory for other recorded verdicts.
- Review the evidence and your use of hippo-memory@1.37.0 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.