code-scanner
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code provided in the
codeparameter to perform security analysis, which could contain instructions intended to influence the agent's behavior. - Ingestion points: The
codeargument in theagent-sec-cli scan-codecommand (SKILL.md). - Boundary markers: The input code is wrapped in single quotes within the shell command execution block, but there are no specific instructions to the agent to disregard natural language commands that might be embedded in the code snippets.
- Capability inventory: Invokes a local command-line tool (
agent-sec-cli) to process the input. - Sanitization: No explicit sanitization or escaping of the code string is described in the skill's instructions before it is passed to the CLI tool.
- [COMMAND_EXECUTION]: The skill utilizes a local command-line interface tool,
agent-sec-cli, to perform its primary function of scanning code. This tool appears to be a vendor-provided resource for security analysis.
Audit Metadata