agent-instructor
Pass
Audited by Gen Agent Trust Hub on Apr 28, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: No override or bypass markers were detected. The skill instructions explicitly direct the agent to preserve existing safety guardrails and constraints during the refinement process.
- [DATA_EXPOSURE_AND_EXFILTRATION]: No sensitive file paths, network operations, or hardcoded credentials were found. The skill operates purely on text provided in the user context.
- [OBFUSCATION]: Analysis for Base64, zero-width characters, homoglyphs, and other encoding techniques returned no results.
- [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill does not perform any package installations or remote script executions. It contains no shell commands or dynamic code generation logic.
- [INDIRECT_PROMPT_INJECTION]: While the skill is designed to process external text (instruction files), it lacks any dangerous capabilities (network, file-write, or shell execution) that could be exploited through malicious content in those files. Output is also safely wrapped in quadruple backticks.
Audit Metadata