registry / @firestitch/standards / 1.0.1

@firestitch/standards@1.0.1

Firestitch engineering standards - routing indexes and rule documents

AI Security Review

scanned 18d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. The install hook mutates a consuming repository’s Claude instruction file. It overwrites that foreign AI-agent control surface and makes the generated files ignored by Git.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

npm install

Impact

Unconsented persistent control of Claude-agent instructions in the consumer project.

Mechanism

postinstall overwrite of repository CLAUDE.md plus .gitignore mutation

Policy narrative

On npm install, bin/postinstall.js walks from the current directory to a Git repository root, renders package-controlled agent instructions, and writes them to the consumer’s CLAUDE.md even when an existing file differs. It also creates CLAUDE.local.md and modifies .gitignore to hide both files. This is an unconsented install-time mutation of a foreign, repository-wide AI-agent instruction surface.

Rationale

Source inspection confirms concrete postinstall takeover of a consumer repository’s CLAUDE.md and persistence-oriented ignore-file mutation. This meets the blocking policy for unconsented install-time mutation of a broad AI-agent control surface.

Evidence

package.jsonbin/postinstall.jsCLAUDE.mdCLAUDE.local.md.gitignore

Decision evidence

public snapshot

AI identified this as a dangerous AI-agent capability at 98.0% confidence with false-positive risk.

@firestitch/standards@1.0.1

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code