repo-harness-assessment
Pass
Audited by Gen Agent Trust Hub on Aug 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns were detected. The skill is designed for static analysis of repository metadata, documentation, and entrypoints.
- [COMMAND_EXECUTION]: The skill uses
rgandfindto search for specific files and patterns within a provided repository root path. These operations are limited to local discovery and do not involve privilege escalation or arbitrary code execution. - [PROMPT_INJECTION]: The skill analyzes untrusted repository data which introduces a surface for indirect prompt injection. 1. Ingestion points:
rgandfindoutputs from the repository root. 2. Boundary markers: Absent. 3. Capability inventory: Local file read operations and directory traversal. 4. Sanitization: Absent. - [CREDENTIALS_UNSAFE]: The skill incorporates defensive instructions that explicitly caution against the collection or exposure of private credentials, API keys, and environment-specific configuration.
Audit Metadata