@companion-ai/feynman@0.3.46
Research-first CLI agent built on Pi and alphaXiv
AI Security Review
scanned 2mo ago · by lpm-firewall-aiThe package has a real, user-triggered agent-runtime patching capability, so it should be warned for lifecycle risk. Its preinstall hook is only a Node-version check, and inspection found no concrete malicious behavior.
Decision evidence
public snapshot- The feynman command runs a local Pi patching script before loading the application.
- The patching script rewrites files in installed Pi runtime packages when its patch functions change their source.
- The runtime patch changes Pi memory handling and its child command invocation using Feynman environment variables.
- The package declares bundled Pi extensions, prompts, and skills, making this an agent-runtime extension lifecycle rather than a normal library.
- The only preinstall hook checks the supported Node.js version and does not write files or make network requests.
- The patching path is activated by running the feynman command, not by npm installation.
- The packaged runtime archive has a recorded SHA-256 digest and an explicit runtime dependency manifest.
- No source evidence established credential harvesting, secret exfiltration, destructive behavior, or an unconsented install-time write to a broad foreign agent configuration.
Source & flagged code
13 flaggedPackage defines install-time lifecycle scripts.
package.jsonView on unpkgSource gates dangerous network, credential, or execution behavior behind CI, host, platform, time, or geo fingerprint checks.
bin/feynman.jsView on unpkg · L25Findings
Affected versions and remediation
This report applies to @companion-ai/feynman@0.3.46.
See all versions of @companion-ai/feynman for other recorded verdicts.
- Review the evidence and your use of @companion-ai/feynman@0.3.46 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.