reset-context-contamination
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFE
Full Analysis
- [NO_CODE]: The skill consists entirely of natural language instructions within Markdown files. No scripts (Python, JavaScript, etc.) or executable binaries are present in the package.
- [PROMPT_INJECTION]: Analysis of instructions like 'Discard the accumulated drafts' and 'Void the earlier drafts entirely' confirmed they are intended for legitimate context management. These directives do not attempt to bypass AI safety filters or extract system prompts.
- [SAFE]: The skill does not perform any network operations, access sensitive files, or require any external credentials. All instructions are transparent and consistent with the stated goal of re-deriving task statements.
Audit Metadata