registry / @pedrofariasx/toolstream / 0.3.0

@pedrofariasx/toolstream@0.3.0

Incremental streaming parser for LLM tool calls — framework-agnostic, event-driven, multi-provider

AI Security Review

scanned 4h ago · by lpm-firewall-ai

LPM blocks this version under the AI-agent control-surface policy. npm postinstall mutates the consuming project's Claude AI configuration area. It creates a directory and writes an auto-generated source file outside the package.

Static reason

One or more suspicious static signals were detected.

Trigger

Installing the package through npm-compatible clients.

Impact

Unconsented modification of a foreign AI-agent control surface.

Mechanism

postinstall writes project .claude/sources/toolstream.md

Policy narrative

On installation, the lifecycle hook uses INIT_CWD or the current directory as the consumer project and recursively creates .claude/sources, then writes toolstream.md there. This modifies an AI-agent-facing project configuration surface without an explicit user command.

Rationale

The postinstall hook concretely performs unconsented mutation of the consuming project's .claude control surface. Although its content is a benign guide and no exfiltration was found, this meets the install-time AI control-surface block boundary.

Evidence

package.jsonscripts/postinstall.mjsai/CLAUDE.md.ai-context/integration.md.claude/sources/toolstream.md

Decision evidence

public snapshot

AI called this Malicious at 97.0% confidence as Malware with false-positive risk.

@pedrofariasx/toolstream@0.3.0

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code