The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill instructions describe the use of administrative privileges (sudo) to perform high-impact system modifications:
Binding a local proxy to privileged ports 80 (HTTP) and 443 (HTTPS).
Modifying the system /etc/hosts file to register .localhost domains via portless hosts sync.
Installing a generated local Certificate Authority (CA) into the system trust store via the portless trust command to enable trusted HTTPS for local development.
[COMMAND_EXECUTION]: The skill facilitates the execution of arbitrary shell commands as sub-processes (e.g., portless <name> <cmd> [args...]), which is used to wrap and proxy development servers.
[EXTERNAL_DOWNLOADS]: The skill guides users to install an external package, portless, from the NPM registry using commands like npm install -g portless.
[PERSISTENCE]: The tool creates persistent configuration changes on the host system, specifically in the /etc/hosts file and the system's root certificate store, which remain active after the skill execution concludes.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect injection surface by automatically inferring project metadata from the local environment to construct command parameters.
Ingestion points: Metadata parsed from package.json, git root directories, and current working directory names.
Boundary markers: No specific boundary markers or instruction-override protections are mentioned for data parsed from local project files during automatic name inference.
Capability inventory: The skill can execute shell commands (portless run), modify system files (/etc/hosts), and modify system security settings (CA trust store).
Sanitization: There is no evidence of sanitization or strict schema validation for the names inferred from the file system before they are incorporated into shell commands.

portless-local