context-engineering
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION] (SAFE): No attempts to bypass safety filters or override agent instructions. The skill includes educational content on 'context poisoning' and adversarial prompts, but this is presented solely for defensive analysis and mitigation purposes.
- [DATA_EXFILTRATION] (SAFE): No hardcoded credentials, sensitive file path access, or network operations (curl, wget, fetch) were found. The skill focuses on internal logic and architectural design rather than external data transmission.
- [REMOTE_CODE_EXECUTION] (SAFE): The skill does not download or execute remote scripts. While it references local Python scripts (context_analyzer.py and compression_evaluator.py) in the SKILL.md metadata, these files are not present in the audited bundle and the documentation describes them as analytical tools.
- [COMMAND_EXECUTION] (SAFE): No dangerous shell commands, privilege escalation (sudo), or persistence mechanisms were identified. Code snippets provided in reference files are illustrative examples of Python classes and functions, not executable system commands.
- [OBFUSCATION] (SAFE): All files consist of plain-text Markdown and clear code examples. No Base64 encoding, zero-width characters, or homoglyph-based evasion techniques were detected.
Audit Metadata