registry / @realvendex/pi-loop / 1.0.2

@realvendex/pi-loop@1.0.2

Pi coding-agent extension that adds a /loop command — run a prompt repeatedly with convergence detection, rich stop conditions, lifecycle hooks, a live TUI panel, and desktop notifications.

AI Security Review

scanned 19h ago · by lpm-firewall-ai

LPM treats this as warn-only first-party agent extension lifecycle risk. A Pi agent extension auto-loads and exposes an explicit /loop command. It can repeatedly send prompts, auto-approve confirmations when --yes is chosen, and run user-specified shell hooks.

Static reason

No blocking static signals were detected.

Trigger

Pi loads the extension, then the user invokes /loop with a prompt and optional flags.

Impact

A user can deliberately grant repeated agent actions and shell execution; no unconsented install-time action or exfiltration is established.

Mechanism

Agent-loop automation with explicit shell-hook execution and local run logging.

Rationale

Source establishes an auto-loaded agent extension with intentional high-impact automation features, but no concrete malicious chain. Warn under the extension/dangerous-capability policy rather than block.

Evidence

package.jsonextensions/index.tsREADME.md~/.pi/loop-state.json~/.pi/loops/loop-*.json

Decision evidence

public snapshot

AI called this Suspicious at 92.0% confidence as Dangerous Capability with low false-positive risk.

Evidence for warning

extensions/index.ts registers a Pi extension through the package manifest.

@realvendex/pi-loop@1.0.2

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code