test
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill identifies and executes arbitrary shell commands found in project configuration files like
package.json,Makefile,Taskfile.yml, and CI workflows. This allows for the execution of scripts defined within the target codebase. - [INDIRECT_PROMPT_INJECTION]: The skill's workflow depends on processing untrusted data from the target codebase, creating a vulnerability surface.
- Ingestion points: Project configuration files (e.g.,
package.json,pyproject.toml,Makefile) and test files. - Boundary markers: The skill does not employ explicit delimiters or safety instructions to isolate untrusted project data from the agent's execution logic.
- Capability inventory: The skill maintains high autonomy to execute shell commands, modify source code, and delete files to resolve test failures.
- Sanitization: Discovered commands are executed without rigorous validation or sandboxing beyond the base execution environment constraints.
Audit Metadata