The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data by reading issue comments and status information from Drupal.org via the drupalorg CLI (SKILL.md §1, §5). This creates a surface for indirect prompt injection. While it lacks explicit prompt boundary markers for the comment content, the skill implements strong countermeasures: it mandates regular expression validation for all identifiers (^[0-9]+$ for issue IDs and ^[a-z][a-z0-9_]*$ for project names) and strictly prohibits interpolating unsanitized input into shell commands or building subcommands from user input (SKILL.md §5). The skill's capabilities include executing git and drupalorg CLI commands.

contribution-issue