security-review
Fail
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: CRITICALINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFESAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code provided by users and has access to high-privilege tools such as
Bash. This creates an attack surface where instructions hidden in the code being audited could attempt to manipulate the agent. * Ingestion points: User-provided code, diffs, or files. * Boundary markers: None explicitly defined. * Capability inventory:Read,Grep,Glob,Bash,Task. * Sanitization: None explicitly defined. - [REMOTE_CODE_EXECUTION]: Automated detections of 'reverse_shell_pattern' and 'eval_with_user_input' are false positives. These snippets are located in documentation files like
references/supply-chain.mdto illustrate malicious patterns for the agent to detect and do not represent executable code within the skill. - [CREDENTIALS_UNSAFE]: References to hardcoded secrets (e.g., 'sk-...') are provided as example patterns for detection and do not expose real credentials.
- [SAFE]: The skill is authored by 'getsentry', a known entity in the security space, and uses well-established security reference material from the OWASP Cheat Sheet Series.
Recommendations
- CRITICAL: 1 file(s) identified as malware by FileRep - DO NOT USE
Audit Metadata