The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/pcp script executes multiple subprocesses, including wp-cli, python3, and php. It passes user-supplied arguments and the current working directory path to these tools as distinct arguments to mitigate shell injection.
[DYNAMIC_EXECUTION]: The skill implements logic for site lookup and JSON parsing through inline Python scripts executed via heredocs. It also dynamically resolves the paths to PHP and MySQL binaries based on the versions specified in Local's configuration files at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests data from the file system and command-line arguments to determine its behavior. The output of the plugin check command, which scans third-party WordPress plugin code, is returned directly to the agent's context.
Ingestion points: Current working directory, command-line arguments, and the contents of sites.json.
Boundary markers: No explicit delimiters or instructions are used to prevent the agent from following directions potentially embedded in the plugin source code or check results.
Capability inventory: The script can execute WP-CLI commands and PHP scripts in the context of the local WordPress site.
Sanitization: While shell injection is mitigated by using bash arrays for execution, the content returned to the agent context is not sanitized or escaped.

wp-pcp-local