AI Security Review
scanned 3h ago · by lpm-firewall-aiNo confirmed malicious attack surface. The package is an AI API bridge whose network calls and optional ACP process launches are tied to explicit CLI/server functionality and configuration.
Decision evidence
public snapshot- `lib/acp-launcher.js` can start ACP commands, including `npx`/`bunx`, but only through runtime server behavior gated by `ACPTOAPI_ENABLE_ACP=1`.
- `bin/acptoapi.js --update` explicitly runs npm cache/version commands; it is not an npm lifecycle hook.
- `lib/acp-launcher.js` appends operational logs under `~/.claude/gm-log`, but does not alter agent configuration or instructions.
- `package.json` has no `preinstall`, `install`, `postinstall`, or other lifecycle scripts.
- `index.js` import path only exports SDK functions; process spawning is confined to explicit `runClaude` usage.
- `bin/acptoapi.js` launches its server only through the user-invoked CLI.
- `install-acp-agents.ps1` performs global installs only when the separate PowerShell script is explicitly run.
- `lib/config.js` parses JSON configuration rather than evaluating configuration code.
- Reviewed provider networking routes requests to configured AI backends; no credential harvesting or unrelated exfiltration was found.
Source & flagged code
6 flagged · loading sourceThis package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
bin/acptoapi.jsView on unpkgPackage source invokes a package manager install command at runtime.
bin/acptoapi.jsView on unpkg · L123A single source file combines environment access, network access, and code or shell execution; review context before blocking.
index.jsView on unpkg · L122Package ships non-JavaScript build or shell helper files.
install-acp-agents.ps1View on unpkg