The Agent Skills Directory

High-Privilege Container Execution: The skill provides instructions to run Docker containers using the --privileged flag, along with specific capabilities like SYS_ADMIN, MKNOD, and direct access to the /dev/fuse device.
Context: These permissions are required to allow a containerized process to interact with the host's kernel FUSE system for debugging.
Security Consideration: Running containers with these flags significantly reduces the isolation between the container and the host system, potentially allowing the containerized process to access host resources beyond the intended scope.
Dynamic Script Generation and Execution: The workflow involves creating a local Node.js environment (package.json) and an execution script (probe.mjs) at runtime, which are then executed via node.
Context: This is used to create a controlled environment to drive RPC calls against the debugging target.
Security Consideration: Automated generation and execution of scripts is a dynamic execution pattern that should be monitored to ensure only the intended code is run.
Indirect Prompt Injection Surface: The skill is designed to process and synchronize data between a local SQLite database and an external containerized service via RPC.
Evidence: The pullOnce and pushOnce operations in probe.mjs move data across process boundaries.
Security Consideration: If the data being processed originated from an untrusted source within the FUSE mount or the container, it could theoretically contain instructions intended to influence the agent's logic. This is a standard risk for skills involving data ingestion.
Secure Dependency Management: The skill explicitly uses the --ignore-scripts flag during npm install.
Context: This is a security best practice that prevents the execution of potentially unsafe lifecycle scripts (like preinstall or postinstall) from third-party packages.

debugging-computerd-fuse