registry / @mounaji_npm/agent-runner / 0.4.0

@mounaji_npm/agent-runner@0.4.0

Portable headless coding-agent runner — polls CognitionDesk for pending claude_code-target loops and executes them via pluggable engines (claude-cli today; agent-sdk later). Extracted and generalized from mcp-mounaji-rag's loop-runner.ts.

AI Security Review

scanned 4h ago · by lpm-firewall-ai

Review flagged AI-agent configuration or capability changes. This remains warn-only unless evidence shows foreign-agent hijack through preinstall/install/postinstall, hidden persistence, exfiltration, remote code execution, or other concrete malicious behavior.

Static reason

No blocking static signals were detected.

Trigger

User runs `agent-runner start` or calls `createAgentRunner().start()` with service credentials.

Impact

A party controlling the configured backend can direct Claude Code to act in the configured working directory and access environment available to that child process.

Mechanism

Poll remote instructions and spawn Claude Code with prompt stdin, cwd, and inherited environment.

Rationale

No covert install-time behavior, hard-coded exfiltration endpoint, or foreign agent-control mutation was found. The documented remote prompt-to-Claude execution path remains a dangerous dual-use capability and merits a warning.

Evidence

package.jsonsrc/index.jssrc/engines/claudeCli.jssrc/prompts.jssrc/heartbeat.jssrc/events.jsbin/agent-runner.js

Network endpoints3

${COGNITIONDESK_URL}/platform-api/loops${NEXT_PUBLIC_SUPABASE_URL}/rest/v1/mcp_instances${NEXT_PUBLIC_SUPABASE_URL}/rest/v1/mcp_events

@mounaji_npm/agent-runner@0.4.0

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code