The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of powerful shell commands via the Bash tool to perform database resets, user deletions, and arbitrary SQL queries. Commands like wp db reset --yes and wp db query allow for direct and destructive manipulation of the system and its data.
[CREDENTIALS_UNSAFE]: The documentation and examples include patterns for passing sensitive credentials, such as database and user passwords, as plaintext command-line arguments (e.g., --dbpass=<password> and --user_pass=<password>). This practice leads to credential exposure in shell history and process monitoring tools.
[EXTERNAL_DOWNLOADS]: The skill documents the capability to install plugins and themes from arbitrary external URLs or ZIP files via wp plugin install <slug|url|zip>, which can be used to download and execute code from untrusted sources.
[PROMPT_INJECTION]: The skill exhibits an indirect injection surface in the 'Content Import from CSV' workflow.
Ingestion points: External CSV files (import.csv, import-meta.csv) are processed in references/examples.md.
Boundary markers: Absent; the shell script iterates through the file content without structural delimiters or warnings to ignore embedded instructions.
Capability inventory: Uses wp post create and wp post meta add to inject data into the WordPress database.
Sanitization: Employs minimal sanitization (sed 's/"//g'), which is insufficient to prevent command injection or data corruption if the input CSV contains malicious payloads tailored for shell interpolation.

wp-cli