The Agent Skills Directory

Unverifiable Dependencies & Remote Code Execution (HIGH): The script scripts/run_portpilot.sh includes the command npx -y portpilot-cli@1 as a fallback. This automatically downloads and executes code from the public NPM registry at runtime, which bypasses local integrity checks and allows for potential Remote Code Execution from an untrusted external source.
Privilege Escalation (HIGH): The SKILL.md and references/safety-policy.md files instruct the agent to proactively request require_escalated permissions and set a reusable prefix_rule for the node command. This 'Permission bootstrap' is designed to persist elevated privileges and suppress future security confirmations for the CLI tool for the remainder of the session.
Data Exposure & Exfiltration (LOW): The skill accesses sensitive local system state, including process IDs (PIDs), command lines, and current working directories of running processes. It also maintains a state directory in the user's home folder (~/.portpilot), involving direct filesystem access.
Dynamic Execution (LOW): The skill dynamically resolves absolute paths to its bundled JavaScript CLI scripts and executes them using the node runtime via shell scripts, which is a common but powerful capability that increases the skill's attack surface.

portpilot-assistant