@accio-ai/cli@0.1.61
Work Agent CLI - AI-native command line interface
AI Security Review
scanned 27d ago · by lpm-firewall-aiThe postinstall hook and binary launcher stay within this CLI: cache cleanup under its own config directory and a local spawn of the matching platform binary. Static inspection found no exfiltration, remote execution, self-install chain, or unconsented agent control-surface write.
Decision evidence
public snapshot- package.json exposes only the workctl bin and a postinstall that runs node install.js. Optional dependencies are same-version platform packages, not a self-dependency on @accio-ai/cli.
- install.js resolves HOME or WORKCTL_CONFIG_DIR and recursively deletes only that directory's cache subdirectory.
- bin/workctl.js resolves a local platform or vendor binary and spawns it with the user's arguments and inherited stdio, after deleting WORKCTL_MCP_URL, WORKCTL_CATALOG_FIXTURE, and WORKCTL_RUNTIME_CONFIG. No network, eval, or credential send is present.
- Postinstall deletes the package cache directory without an extra user prompt, but the path is limited to the CLI's own config cache and does not write agent control files.
Source & flagged code
6 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.json exposes only the workctl bin and a postinstall that runs node install.js. Optional dependencies are same-version platform packages, not a self-dependency on @accio-ai/cli.
package.jsonView on unpkg · L14Findings
Affected versions and remediation
This report applies to @accio-ai/cli@0.1.61.
See all versions of @accio-ai/cli for other recorded verdicts.
- Review the evidence and your use of @accio-ai/cli@0.1.61 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.