npm AI-agent policy reports, with evidence.

The confirmed issue is not generic network or credential handling; it is install-time mutation of Claude skills and MCP configuration, an AI-agent control surface. Under the firewall boun...

Source inspection confirms unconsented lifecycle writes to Claude Code skill directories, which is a concrete AI-agent control-surface mutation even though the content matches the package...

Source inspection confirms a postinstall hook that writes package-controlled Claude skill instructions into the user's AI agent configuration area without an explicit setup command. That...

Although the bundled skills look aligned with the CLI, writing package-controlled AI-agent skills during postinstall is an unconsented lifecycle mutation of agent control files. No exfilt...

Static inspection confirms install-time mutation of Cursor/Claude/Codex agent control files and hook execution surfaces, which is not a normal passive library install behavior. The behavi...

Direct source inspection confirms install-time writes into Claude's skills directory with autonomous agent instructions, which is a concrete AI-agent control-surface mutation outside norm...

The suspicious behavior is not merely a wrapper or package-aligned CLI capability; the lifecycle hook automatically invokes an agent-skill installer during npm install. Under the firewall...

Although many capabilities are package-aligned for an AI workflow CLI, installing AI-agent control files at lifecycle time is an unconsented control-surface mutation. No exfiltration was...

Source inspection confirms install-time writes of AI-agent rule files into the consumer project root. Even though the content is package-aligned and no traditional malware behavior was fo...

Although the package is otherwise transparent and lacks exfiltration or destructive behavior, the global postinstall hook performs an unconsented lifecycle write into Claude Code's skill...

Direct inspection confirms concrete AI-agent control-surface mutation: the installer writes Claude settings that bypass permission checks and expands trusted token-provisioning behavior....

Static inspection confirms unconsented lifecycle mutation of a host AI-agent control surface via postinstall, which is blocking behavior for the firewall even without credential theft or...

Static source inspection confirms unconsented lifecycle mutation of Codex/Claude agent skill and hook configuration, which is a concrete publish-block condition even though the CLI behavi...

The package has substantial benign, user-invoked media/OSINT functionality, but the postinstall script performs a concrete unconsented mutation of an AI-agent dependency config file. That...

Static inspection confirms install-time writes to OpenCode control-surface configs; under the firewall policy this is blocking behavior even though the implementation otherwise appears pa...

The package does not show credential theft, exfiltration, persistence, or destructive behavior, and scanner blob findings are package-aligned skill assets. However, the npm lifecycle hook...

Static inspection confirms concrete runtime mutation of Claude Code permission settings to bypass tool checks and authorize credential provisioning; this is not merely scanner noise or in...

Static inspection confirms unconsented lifecycle AI-agent control-surface mutation via postinstall writes to .claude, matching the firewall block boundary. No credential theft or network...

Source inspection confirms an install-time write into ~/.claude/skills from package-supplied AI-agent instructions, which is concrete unconsented lifecycle control-surface mutation. Other...

Direct source inspection confirms install-time writes to ~/.claude/skills, which is an AI-agent control surface, from a package lifecycle hook. Even without credential theft or exfiltrati...

Static inspection confirms concrete install-time writes of AI-agent instruction files into consumer projects. No exfiltration was found, but this matches the firewall policy for unconsent...