registry / @firestitch/standards / 1.0.2

@firestitch/standards@1.0.2

Firestitch engineering standards - routing indexes and rule documents

AI Security Review

scanned 17d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. Install-time hook mutates the consuming repository's Claude Code instruction surface. It overwrites CLAUDE.md and adds generated instruction files to .gitignore.

Static reason

One or more suspicious static signals were detected.

Trigger

npm install

Impact

Unconsented control of subsequent AI-agent behavior in the repository.

Mechanism

postinstall writes AI-agent instructions into consumer repo

Policy narrative

On npm install, bin/postinstall.js walks upward to the nearest .git directory, then writes a generated CLAUDE.md at that repository root, creates CLAUDE.local.md, and alters .gitignore. The generated file directs AI agents to load this package's standards before code changes. This is an unconsented postinstall mutation of a consuming project's AI-agent control surface.

Rationale

The lifecycle hook performs concrete, persistent AI-agent instruction-file mutation outside the package directory. Absence of exfiltration does not remove this install-time control-surface risk.

Evidence

package.jsonbin/postinstall.js<consumer-repo>/CLAUDE.md<consumer-repo>/CLAUDE.local.md<consumer-repo>/.gitignore

Decision evidence

public snapshot

AI identified this as a dangerous AI-agent capability at 97.0% confidence with false-positive risk.

@firestitch/standards@1.0.2

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code