harness-init
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill provides standard operational instructions for project initialization and maintenance. No attempts to override system safety guidelines or use jailbreak techniques were detected.
- [DATA_EXFILTRATION]: There are no commands or instructions that attempt to access sensitive system files (e.g., SSH keys, cloud credentials) or exfiltrate data to external servers.
- [REMOTE_CODE_EXECUTION]: The skill does not contain any patterns for downloading and executing code from external URLs or untrusted sources.
- [OBFUSCATION]: No evidence of encoded strings (Base64, hex), zero-width characters, or homoglyph-based obfuscation was found in the instructions or templates.
- [INDIRECT_PROMPT_INJECTION]: The project structure encourages versioning documentation within the repository. While this involves the agent reading and writing files, it does not involve processing untrusted external inputs like user-provided web content, minimizing injection risks.
Audit Metadata