cwe-injection-attacks
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [Indirect Prompt Injection Surface]: This skill processes untrusted application source code to detect vulnerabilities, which is an inherent part of its security auditing function.
- Ingestion points: The skill systematically scans application source files (e.g., .java, .py, .cs, .js, .ts) provided in the workspace (SKILL.md).
- Boundary markers: The skill does not explicitly instruct the agent to ignore potentially malicious instructions embedded in comments or strings within the code being analyzed.
- Capability inventory: The skill is designed to produce a structured JSON report and interacts with a
record-resultcontract to log findings. - Sanitization: The skill does not specify sanitization or escaping mechanisms for the data extracted from source code before it is included in the output report.
Audit Metadata