registry / @vib795/agent-memory / 0.1.0

@vib795/agent-memory@0.1.0

Durable cross-repo knowledge graph for GitHub Copilot and Claude Code. Markdown source of truth, disposable SQLite index, zero runtime dependencies.

AI Security Review

scanned 7d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. Install-time code mutates user-global instruction surfaces for multiple AI agents. It installs three package-provided skills without an explicit user command.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

npm install (postinstall)

Impact

Future Copilot, Claude Code, Codex, or VS Code agent sessions may load package-controlled instructions.

Mechanism

automatic global AI-agent skill linking and prompt-file writes

Policy narrative

On npm installation, the postinstall hook calls setup. Setup writes or symlinks the package’s skill instructions into user-global directories for several unrelated AI-agent products and detected VS Code installations, so later agent sessions can automatically consume package-controlled instructions.

Rationale

The source confirms an unconsented postinstall mutation of broad, foreign AI-agent control surfaces. No separate exfiltration or payload execution was found, but the install-time control-surface write meets the blocking boundary.

Evidence

package.jsonscripts/postinstall.jssrc/setup.jssrc/targets.jssrc/config.js~/.agents/skills/{handoff,recall,remember}~/.claude/skills/{handoff,recall,remember}~/.codex/skills/{handoff,recall,remember}VS Code User/prompts/{handoff,recall,remember}.prompt.md~/.agents/memory

@vib795/agent-memory@0.1.0

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code