banana-pro-director
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a set of sophisticated prompt engineering instructions designed to guide an AI agent in generating text-based image prompts for third-party platforms.
- [PROMPT_INJECTION]: No malicious override or bypass attempts were found. The use of instructional terms like 'CRITICAL' and 'IMPORTANT' is focused on maintaining workflow integrity and output quality.
- [DATA_EXPOSURE_EXFILTRATION]: The skill does not access sensitive local files or perform any network operations to external domains. It focuses entirely on processing user-provided descriptions and visual references.
- [REMOTE_CODE_EXECUTION]: There are no patterns involving the download or execution of remote scripts or the installation of third-party packages.
- [OBFUSCATION]: All content is in clear text. No evidence of Base64 encoding, zero-width characters, or other techniques to hide intent was detected.
- [INDIRECT_PROMPT_INJECTION]: While the skill ingests untrusted user descriptions and 'studies' visual references, it includes strong mitigation patterns: it requires the agent to mirror back a 'locked spec' for human approval before generating final output, and it explicitly forbids inventing details not present in references.
Audit Metadata