registry / @markjaquith/agency / 2.71.12

@markjaquith/agency@2.71.12

Manage agentic work across repositories with durable workbases

AI Security Review

scanned 15d ago · by lpm-firewall-ai

LPM treats this as warn-only first-party agent extension lifecycle risk. The manifest declares postinstall setup for a globally loaded Pi extension. The shipped extension augments Pi prompts only after it detects an Agency workbase; the installer implementation is absent.

Static reason

One or more suspicious static signals were detected.

Trigger

npm installation; then a Pi agent starts inside an Agency workbase.

Impact

Can influence Pi agent context and expose configured checkout skill paths within Agency workbases; no exfiltration or remote payload chain is established.

Mechanism

global Pi extension registration and workbase-scoped prompt/resource injection

Rationale

This is not concrete malware, but an install-time global AI-agent extension is a meaningful control-surface risk. Downgrade to warn under the extension lifecycle policy.

Evidence

package.jsonREADME.mdpi-extensions/agency.ts~/.pi/agent/extensions/agency.ts

Decision evidence

public snapshot

AI called this Suspicious at 90.0% confidence as Dangerous Capability with low false-positive risk.

Evidence for warning

postinstall invokes a Pi-extension installer.
Documentation says the extension is installed globally without project trust.

@markjaquith/agency@2.71.12

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code