quarkus-module-build
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses user-supplied module names (from
$ARGUMENTS) to construct and execute shell commands such asfind,cd, andmvn. If the agent platform's execution environment does not perform automatic sanitization, an attacker could provide module names containing shell metacharacters (e.g.,;,&,|,`) to execute arbitrary commands on the system. - [INDIRECT_PROMPT_INJECTION]: The skill accepts untrusted user input and incorporates it into instructions for subagents without sufficient validation or boundary markers.
- Ingestion points: Module names provided by the user via the
$ARGUMENTSvariable inSKILL.md. - Boundary markers: None. Module paths and names are directly interpolated into the subagent prompt templates without delimiters or "ignore instructions" warnings.
- Capability inventory: The skill performs file system searches, directory changes, and executes the Maven build tool (
mvn). - Sanitization: The instruction for parsing module names only handles space-delimited strings and filters out the word "and," which does not protect against directory traversal or shell injection patterns.
Audit Metadata