drupal-update
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill fetches and processes external content from Drupal module release pages, creating a surface for indirect prompt injection.
- Ingestion points: The script
scripts/fetch_changelog.phpfetches HTML content fromhttps://www.drupal.org/project/[module]/releases/[version]. - Boundary markers: The skill instructions do not specify boundary markers or instructions to ignore potential directives embedded within the fetched changelog text.
- Capability inventory: The agent has extensive shell capabilities including
composer,drush, andddevcommand execution, as well as file writing. - Sanitization: While the PHP script limits string lengths and filters for specific keywords, it does not sanitize the content for LLM safety or escape potential command-like sequences.
- [EXTERNAL_DOWNLOADS]: The skill retrieves information from a remote source at runtime.
- The
scripts/fetch_changelog.phpscript performs GET requests todrupal.orgto fetch release notes. Asdrupal.orgis a well-known and trusted service within the Drupal ecosystem, this is considered a safe and standard operation for its intended purpose. - [COMMAND_EXECUTION]: The skill relies on executing a series of shell commands to manage the Drupal environment.
- The agent is instructed to run
composer update,composer require,drush updb,drush cex, andddev snapshotvia theddevCLI tool. These operations are core to the skill's functionality and are performed within the scoped DDEV container environment.
Audit Metadata