cms-framework-attack
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill framework presents a significant attack surface for indirect prompt injection as it is designed to ingest and process data from external, untrusted web targets.
- Ingestion points: The skill utilizes
WebFetchto retrieve content from target websites for fingerprinting and version identification (e.g.,curlcommands inpayloads.mdandguides/*.md). It also processes output from specialized scanners likewpscan,joomscan, andnucleito determine subsequent attack steps. - Boundary markers: The instructions lack delimiters or explicit warnings (e.g., "ignore embedded instructions") when handling data retrieved from external sources.
- Capability inventory: The skill possesses powerful capabilities including
Bashfor command execution,WebFetchfor network operations, andWritefor file modifications, which could be abused if an attacker-controlled target provides malicious instructions. - Sanitization: No logic is defined for sanitizing, escaping, or validating content retrieved from external targets before it is used to influence the agent's decision-making or command generation.
- [COMMAND_EXECUTION]: The skill methodology relies extensively on the
Bashtool to execute a wide variety of security tools and exploitation scripts. - Evidence: The guides and payload collection provide command templates for running
wpscan,joomscan,droopescan,cmseek,nikto,whatweb, andnuclei. - Third-party scripts: The skill methodology includes the execution of specific exploitation scripts such as
ruby drupalgeddon2.rbandpython3 joomla_rce.py(referenced inguides/joomla-drupal-cms-attack-guide.mdandpayloads.md). These are external scripts whose source and integrity are not verified within the skill's instructions. - [DATA_EXFILTRATION]: A core objective of the skill is the systematic discovery and extraction of sensitive configuration files and database credentials from targets.
- Sensitive paths: The methodology includes targeted searching for files such as
wp-config.php,configuration.php,settings.php, and their backup variants (e.g.,.bak,.old,~), as well as.envfiles and database dumps (backup.sql,db.sql). - Credential extraction: The skill provides specific commands (e.g.,
grep -E "DB_NAME|DB_USER|DB_PASSWORD|DB_HOST" wp-config.php) to extract plaintext credentials for database access, which is an inherent part of the documented attack chain.
Audit Metadata